[refactor] modularize Slint video rendering backends

src/gui/application/gui_application.cpp

@@ -1,7 +1,6 @@
 #include "application/gui_application.h"
 
 #include <SDL3/SDL.h>
-#include <libyuv.h>
 #include <slint.h>
 #include <tinyfiledialogs.h>
 
@@ -29,16 +28,14 @@
 #include "crossdesk_ui.h"
 #include "fa_solid_900.h"
 #include "localization.h"
-#include "nv12_scaler.h"
 #include "platform.h"
+#include "platform/video_renderer.h"
+#include "rendering/slint_video_presenter.h"
 #if _WIN32
 #include <windows.h>
-#include <GL/gl.h>
 
 #include "platform/tray/win_tray.h"
-#include "platform/windows_opengl_video_renderer.h"
 #elif defined(__APPLE__)
-#include "platform/metal_video_renderer.h"
 #include "platform/tray/mac_tray.h"
 #include "platform/window_drag.h"
 #elif defined(__linux__)
@@ -60,41 +57,6 @@ namespace {
 
 using namespace std::chrono_literals;
 
-#if !defined(__APPLE__)
-constexpr GLint kGlClampToEdge = 0x812F;
-#else
-constexpr int kMetalAttachmentAttemptLimit = 30;
-#endif
-
-struct VideoRenderSize {
-  int width = 0;
-  int height = 0;
-
-  bool operator==(const VideoRenderSize&) const = default;
-};
-
-VideoRenderSize FitVideoToRenderArea(int source_width, int source_height,
-                                     int area_width, int area_height) {
-  if (source_width <= 0 || source_height <= 0 || area_width <= 0 ||
-      area_height <= 0) {
-    return {};
-  }
-
-  const double scale =
-      std::min({1.0, static_cast<double>(area_width) / source_width,
-                static_cast<double>(area_height) / source_height});
-  int width = std::max(2, static_cast<int>(std::floor(source_width * scale)));
-  int height =
-      std::max(2, static_cast<int>(std::floor(source_height * scale)));
-
-  // NV12 chroma samples cover 2x2 pixels. Keep the render buffer even-sized
-  // so libyuv never needs to read a partial chroma sample at the edge.
-  width &= ~1;
-  height &= ~1;
-  return {std::min(width, source_width & ~1),
-          std::min(height, source_height & ~1)};
-}
-
 std::string CreatePasswordChangeRequestId(uint64_t sequence) {
   const auto timestamp = std::chrono::system_clock::now()
                              .time_since_epoch()
@@ -930,19 +892,7 @@ struct GuiApplication::SlintUi {
   std::shared_ptr<slint::VectorModel<slint::SharedString>>
       controller_name_model =
           std::make_shared<slint::VectorModel<slint::SharedString>>();
-  std::unordered_map<std::string, uint64_t> displayed_frame_sequence;
-  std::vector<uint8_t> scaled_video_frame;
-  std::vector<uint8_t> scaled_video_frame_scratch;
-#if !defined(__APPLE__)
-  std::mutex video_gl_mutex;
-  std::vector<uint8_t> video_gl_conversion_frame;
-  std::vector<uint8_t> video_gl_pending_frame;
-  uint32_t video_gl_texture = 0;
-  VideoRenderSize video_gl_texture_size;
-  VideoRenderSize video_gl_image_size;
-  VideoRenderSize video_gl_pending_size;
-  bool video_gl_pending_dirty = false;
-#endif
+  std::unique_ptr<SlintVideoPresenter> video_presenter;
   std::vector<std::string> tab_order;
   std::vector<std::string> tab_ids;
   std::string tab_model_signature;
@@ -957,9 +907,6 @@ struct GuiApplication::SlintUi {
   WindowDragState server_drag;
   int main_native_titlebar_attempts = 30;
   int stream_live_resize_configuration_attempts = 0;
-#if defined(__APPLE__)
-  int stream_video_attachment_attempts = 0;
-#endif
   int stream_initial_position_attempts = 0;
   int server_initial_position_attempts = 0;
 #if _WIN32
@@ -1114,15 +1061,9 @@ void GuiApplication::InitializeModules() {
   if (modules_inited_) {
     return;
   }
-#if defined(_WIN32)
-  windows_opengl_video_renderer_ =
-      std::make_unique<WindowsOpenGlVideoRenderer>();
-#elif defined(__APPLE__)
-  // Metal setup can log device, shader, or pipeline failures. Initialize it
-  // here, after Run() has selected the application's log directory, rather
-  // than from GuiRuntime's constructor before InitializeLogger().
-  mac_metal_video_renderer_ = std::make_unique<MacMetalVideoRenderer>();
-#endif
+  // Renderer construction can log device, shader, or pipeline failures.
+  // Initialize it after Run() has selected the application's log directory.
+  video_renderer_ = CreateVideoRenderer();
   devices_.Initialize();
   CreateConnectionPeer();
   modules_inited_ = true;
@@ -1157,6 +1098,8 @@ void GuiApplication::InitializeUi() {
   }
 #endif
   ui_ = std::make_unique<SlintUi>();
+  ui_->video_presenter =
+      std::make_unique<SlintVideoPresenter>(*video_renderer_);
 #if _WIN32
   ui_->main->set_custom_titlebar(true);
 #elif defined(__linux__)
@@ -2261,9 +2204,6 @@ void GuiApplication::SyncStreamKeyboardFocus() {
 }
 
 void GuiApplication::SyncStreamWindow() {
-#if defined(__APPLE__)
-  bool stream_window_created_this_sync = false;
-#endif
   bool has_sessions = false;
   {
     std::shared_lock lock(remote_sessions_mutex_);
@@ -2275,15 +2215,7 @@ void GuiApplication::SyncStreamWindow() {
     (*ui_->stream)->set_custom_titlebar(use_xwayland_gui_);
 #endif
     RegisterFontAwesome((*ui_->stream)->window());
-#if defined(__APPLE__)
-    // Commit the first visible AppKit window as an ordinary opaque window.
-    // The process-wide Slint event loop is already running here, so
-    // appkit_view() may exist before show(); that does not mean the native
-    // titlebar style has been committed yet.
-    (*ui_->stream)->set_native_video_enabled(false);
-#else
     ConfigureStreamVideoRenderer();
-#endif
     // Slint globals belong to a component tree. Force the next localization
     // pass to initialize the newly-created, independent stream window tree.
     ui_->localized_language = -1;
@@ -2293,15 +2225,11 @@ void GuiApplication::SyncStreamWindow() {
     (*ui_->stream)->set_stats_rows(ui_->stats_model);
     BindStreamCallbacks();
     (*ui_->stream)->show();
-#if defined(__APPLE__)
-    stream_window_created_this_sync = true;
-#endif
     // Repeat after the next layout pass because the backend resolves the
     // initial DPI and preferred size asynchronously on some platforms.
     ui_->stream_initial_position_attempts = 2;
 #if defined(__APPLE__)
     ui_->stream_live_resize_configuration_attempts = 30;
-    ui_->stream_video_attachment_attempts = kMetalAttachmentAttemptLimit;
 #endif
     stream_window_created_ = true;
     stream_window_inited_ = true;
@@ -2321,24 +2249,10 @@ void GuiApplication::SyncStreamWindow() {
       --ui_->stream_live_resize_configuration_attempts;
     }
   }
-  if (ui_->stream_video_attachment_attempts > 0 &&
-      !stream_window_created_this_sync) {
-    ConfigureStreamVideoRenderer();
-    const bool metal_unavailable =
-        !mac_metal_video_renderer_ || !mac_metal_video_renderer_->IsReady();
-    const bool metal_attached =
-        mac_metal_video_renderer_ && mac_metal_video_renderer_->IsAttached();
-    if (metal_unavailable) {
-      ui_->stream_video_attachment_attempts = 0;
-    } else if (metal_attached) {
-      video_frame_dirty_.store(true, std::memory_order_release);
-      ui_->stream_video_attachment_attempts = 0;
-    } else if (--ui_->stream_video_attachment_attempts == 0) {
-      (*ui_->stream)->set_native_video_enabled(false);
-      LOG_WARN("Unable to attach the Metal video surface; using CPU rendering");
-    }
-  }
 #endif
+  if (ui_->video_presenter && ui_->video_presenter->EnsureAttached()) {
+    video_frame_dirty_.store(true, std::memory_order_release);
+  }
   if (ui_->stream_initial_position_attempts > 0) {
     if (PositionWindowAtCenter(
             (*ui_->stream)->window(), ui_->main->window(),
@@ -2351,29 +2265,20 @@ void GuiApplication::SyncStreamWindow() {
   }
   if (!has_sessions) {
     SetStreamKeyboardFocus(false);
-#if defined(_WIN32)
-    if (windows_opengl_video_renderer_) {
-      windows_opengl_video_renderer_->SetSelectedStream({});
-    }
-#endif
 #if defined(__APPLE__)
     void* stream_view = (*ui_->stream)->window().appkit_view();
     SetStreamWindowFullscreen(false);
     UnregisterStreamWindow(stream_view);
-    if (mac_metal_video_renderer_) {
-      mac_metal_video_renderer_->SetSelectedStream({});
-      mac_metal_video_renderer_->Detach();
-    }
 #endif
+    if (ui_->video_presenter) {
+      ui_->video_presenter->Detach();
+    }
     (*ui_->stream)->hide();
     ui_->stream.reset();
     stream_window_created_ = false;
     stream_window_inited_ = false;
     ui_->stream_initial_position_attempts = 0;
     ui_->stream_live_resize_configuration_attempts = 0;
-#if defined(__APPLE__)
-    ui_->stream_video_attachment_attempts = 0;
-#endif
     focused_remote_id_.clear();
     controlled_remote_id_.clear();
     ui_->tab_order.clear();
@@ -2590,7 +2495,7 @@ void GuiApplication::SyncStreamWindow() {
 }
 
 void GuiApplication::SyncStreamVideoFrame() {
-  if (!ui_ || !ui_->stream) {
+  if (!ui_ || !ui_->stream || !ui_->video_presenter) {
     return;
   }
   auto props = SelectedSession();
@@ -2598,421 +2503,46 @@ void GuiApplication::SyncStreamVideoFrame() {
     return;
   }
 
-#if defined(__APPLE__)
-  if (mac_metal_video_renderer_ && mac_metal_video_renderer_->IsReady() &&
-      mac_metal_video_renderer_->IsAttached()) {
-    const double top_inset =
-        !fullscreen_button_pressed_ && ui_->tab_ids.size() > 1 ? 30.0 : 0.0;
-    const auto render_outcome = mac_metal_video_renderer_->RenderLatest(
-        props->remote_id_, top_inset, !fullscreen_button_pressed_);
-    if (render_outcome.result ==
-        MacMetalVideoRenderer::RenderResult::rendered) {
-      if (!(*ui_->stream)->get_has_frame()) {
-        (*ui_->stream)->set_has_frame(true);
-      }
-      if (!(*ui_->stream)->get_receiving_text().empty()) {
-        (*ui_->stream)->set_receiving_text("");
-      }
-      if (render_outcome.width > 0 && render_outcome.height > 0 &&
-          render_outcome.sequence > 0) {
-        (*ui_->stream)
-            ->set_stats_resolution(
-                UiText(std::to_string(render_outcome.width) + "x" +
-                       std::to_string(render_outcome.height)));
-        if (ui_->displayed_frame_sequence[props->remote_id_] !=
-            render_outcome.sequence) {
-          ui_->displayed_frame_sequence[props->remote_id_] =
-              render_outcome.sequence;
-          ++props->frame_count_;
-        }
-      }
-      return;
-    }
-    if (render_outcome.result == MacMetalVideoRenderer::RenderResult::empty) {
-      if ((*ui_->stream)->get_has_frame()) {
-        (*ui_->stream)->set_has_frame(false);
-      }
-      return;
-    }
-    if (render_outcome.result == MacMetalVideoRenderer::RenderResult::idle) {
-      return;
-    }
-    // A transient Metal error must not change renderer ownership mid-window.
-    // Keep the last frame and retry on the next UI tick.
-    return;
-  }
-#endif
-
-#if defined(_WIN32)
-  if (windows_opengl_video_renderer_ &&
-      windows_opengl_video_renderer_->IsReady()) {
-    windows_opengl_video_renderer_->SetSelectedStream(props->remote_id_);
-    SubmitCachedFrameToOpenGl(props);
-    if (!(*ui_->stream)->get_native_video_enabled()) {
-      (*ui_->stream)->set_native_video_enabled(true);
-    }
-
-    int width = 0;
-    int height = 0;
-    uint64_t sequence = 0;
-    {
-      std::lock_guard lock(props->video_frame_mutex_);
-      width = props->video_width_;
-      height = props->video_height_;
-      sequence = props->video_frame_sequence_;
-    }
-    const bool has_frame = width > 0 && height > 0 && sequence > 0;
-    (*ui_->stream)->set_has_frame(has_frame);
-    if (has_frame) {
-      (*ui_->stream)->set_receiving_text("");
-      (*ui_->stream)
-          ->set_stats_resolution(UiText(std::to_string(width) + "x" +
-                                        std::to_string(height)));
-      if (ui_->displayed_frame_sequence[props->remote_id_] != sequence) {
-        ui_->displayed_frame_sequence[props->remote_id_] = sequence;
-        ++props->frame_count_;
-      }
-    }
-    (*ui_->stream)->window().request_redraw();
-    return;
-  }
-  if ((*ui_->stream)->get_native_video_enabled()) {
-    (*ui_->stream)->set_native_video_enabled(false);
-  }
-#endif
-
-  std::shared_ptr<std::vector<unsigned char>> frame;
-  int width = 0;
-  int height = 0;
-  uint64_t sequence = 0;
+  SlintVideoPresenter::Frame frame;
+  frame.remote_id = props->remote_id_;
   {
     std::lock_guard lock(props->video_frame_mutex_);
-    frame = props->front_frame_;
-    width = props->video_width_;
-    height = props->video_height_;
-    sequence = props->video_frame_sequence_;
+    frame.nv12 = props->front_frame_;
+    frame.width = props->video_width_;
+    frame.height = props->video_height_;
+    frame.sequence = props->video_frame_sequence_;
   }
 
-  const size_t nv12_size = static_cast<size_t>(width) * height * 3 / 2;
-  if (!frame || width <= 0 || height <= 0 || frame->size() < nv12_size) {
-    (*ui_->stream)->set_has_frame(false);
-    return;
+  const auto result = ui_->video_presenter->Present(frame);
+  if (result.width > 0 && result.height > 0) {
+    (*ui_->stream)
+        ->set_stats_resolution(UiText(std::to_string(result.width) + "x" +
+                                      std::to_string(result.height)));
   }
-  if (ui_->displayed_frame_sequence[props->remote_id_] == sequence) {
-    return;
-  }
-
-  const auto window_size = (*ui_->stream)->window().size();
-  const VideoRenderSize render_size = FitVideoToRenderArea(
-      width, height, static_cast<int>(window_size.width),
-      static_cast<int>(window_size.height));
-  const uint8_t* y_plane = frame->data();
-  const uint8_t* uv_plane =
-      frame->data() + static_cast<size_t>(width) * height;
-  int output_width = width;
-  int output_height = height;
-
-  if (render_size.width > 0 && render_size.height > 0 &&
-      (render_size.width != width || render_size.height != height)) {
-    const size_t scaled_nv12_size =
-        static_cast<size_t>(render_size.width) * render_size.height * 3 / 2;
-    ui_->scaled_video_frame.resize(scaled_nv12_size);
-    uint8_t* scaled_y = ui_->scaled_video_frame.data();
-    uint8_t* scaled_uv =
-        scaled_y +
-        static_cast<size_t>(render_size.width) * render_size.height;
-    if (ScaleNv12ViaI420(
-            y_plane, width, uv_plane, width, width, height, scaled_y,
-            render_size.width, scaled_uv, render_size.width, render_size.width,
-            render_size.height, libyuv::kFilterBox,
-            &ui_->scaled_video_frame_scratch) == 0) {
-      y_plane = scaled_y;
-      uv_plane = scaled_uv;
-      output_width = render_size.width;
-      output_height = render_size.height;
-    }
-  }
-
-#if !defined(__APPLE__)
-  uint32_t texture_id = 0;
-  {
-    std::lock_guard lock(ui_->video_gl_mutex);
-    texture_id = ui_->video_gl_texture;
-  }
-  if (texture_id != 0) {
-    ui_->video_gl_conversion_frame.resize(
-        static_cast<size_t>(output_width) * output_height * 4);
-    if (libyuv::NV12ToABGR(
-            y_plane, output_width, uv_plane, output_width,
-            ui_->video_gl_conversion_frame.data(), output_width * 4,
-            output_width, output_height) != 0) {
-      return;
-    }
-    {
-      std::lock_guard lock(ui_->video_gl_mutex);
-      if (ui_->video_gl_texture == 0) {
-        return;
-      }
-      texture_id = ui_->video_gl_texture;
-      ui_->video_gl_pending_frame.swap(ui_->video_gl_conversion_frame);
-      ui_->video_gl_pending_size = {output_width, output_height};
-      ui_->video_gl_pending_dirty = true;
-    }
-    const VideoRenderSize output_size{output_width, output_height};
-    if (ui_->video_gl_image_size != output_size) {
-      (*ui_->stream)
-          ->set_frame(slint::Image::create_from_borrowed_gl_2d_rgba_texture(
-              texture_id,
-              slint::Size<uint32_t>{static_cast<uint32_t>(output_width),
-                                    static_cast<uint32_t>(output_height)}));
-      ui_->video_gl_image_size = output_size;
-    }
-    (*ui_->stream)->set_has_frame(true);
-    (*ui_->stream)->set_receiving_text("");
-    (*ui_->stream)->window().request_redraw();
-    ui_->displayed_frame_sequence[props->remote_id_] = sequence;
+  if (result.new_frame) {
     ++props->frame_count_;
-    return;
   }
-#endif
-
-  slint::SharedPixelBuffer<slint::Rgb8Pixel> pixels(output_width,
-                                                    output_height);
-  if (libyuv::NV12ToRAW(
-          y_plane, output_width, uv_plane, output_width,
-          reinterpret_cast<uint8_t*>(pixels.begin()), output_width * 3,
-          output_width, output_height) != 0) {
-    return;
-  }
-
-  (*ui_->stream)->set_frame(slint::Image(std::move(pixels)));
-  (*ui_->stream)->set_has_frame(true);
-  (*ui_->stream)->set_receiving_text("");
-  ui_->displayed_frame_sequence[props->remote_id_] = sequence;
-  ++props->frame_count_;
 }
 
 void GuiApplication::ConfigureStreamVideoRenderer() {
-#if defined(__APPLE__)
-  if (!ui_ || !ui_->stream) {
+  if (!ui_ || !ui_->stream || !ui_->video_presenter) {
     return;
   }
-  const bool native_video_enabled =
-      mac_metal_video_renderer_ && mac_metal_video_renderer_->IsReady();
-  if (!native_video_enabled) {
-    (*ui_->stream)->set_native_video_enabled(false);
-    return;
-  }
-  if (void* view = (*ui_->stream)->window().appkit_view()) {
-    // Slint applies transparency through a property binding. Set it before
-    // restoring the AppKit titlebar so the later operation wins regardless of
-    // whether the backend commits the binding synchronously or next tick.
-    (*ui_->stream)->set_native_video_enabled(true);
-    if (!mac_metal_video_renderer_->Attach(view)) {
-      (*ui_->stream)->set_native_video_enabled(false);
-    } else if (auto selected_session = SelectedSession()) {
-      // Metal owns a separate monotonic sequence, so discard a CPU-renderer
-      // sequence recorded during the short window-attachment gap.
-      ui_->displayed_frame_sequence.erase(selected_session->remote_id_);
-      if (mac_metal_video_renderer_->SetSelectedStream(
-              selected_session->remote_id_)) {
-        video_frame_dirty_.store(true, std::memory_order_release);
-      }
-      SubmitCachedFrameToMetal(selected_session);
-    }
-  } else {
-    // Keep the initial window opaque until winit has created its native
-    // NSWindow. Creating it transparent opts into a full-size titlebar whose
-    // title field sits underneath the standard traffic-light buttons.
-    (*ui_->stream)->set_native_video_enabled(false);
-  }
-  return;
-#else
-  if (!ui_ || !ui_->stream) {
-    return;
-  }
-  const auto error = (*ui_->stream)->window().set_rendering_notifier(
-      [this](slint::RenderingState state, slint::GraphicsAPI graphics_api) {
-        if (!ui_ || graphics_api != slint::GraphicsAPI::NativeOpenGL) {
-          return;
-        }
 
-        std::lock_guard lock(ui_->video_gl_mutex);
-        if (state == slint::RenderingState::RenderingSetup) {
-          GLuint texture = 0;
-          glGenTextures(1, &texture);
-          if (texture == 0) {
-            return;
-          }
-          GLint previous_texture = 0;
-          glGetIntegerv(GL_TEXTURE_BINDING_2D, &previous_texture);
-          glBindTexture(GL_TEXTURE_2D, texture);
-          glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
-          glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
-          glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, kGlClampToEdge);
-          glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, kGlClampToEdge);
-          glBindTexture(GL_TEXTURE_2D,
-                        static_cast<GLuint>(previous_texture));
-          ui_->video_gl_texture = texture;
-#if defined(_WIN32)
-          if (windows_opengl_video_renderer_ &&
-              windows_opengl_video_renderer_->Setup()) {
-            windows_opengl_video_renderer_->SetSelectedStream(
-                focused_remote_id_);
-            video_frame_dirty_.store(true, std::memory_order_release);
-          }
-#endif
-          return;
+  ui_->video_presenter->PrepareWindow(
+      *ui_->stream,
+      [this] {
+        SlintVideoPresenter::SurfaceState state;
+        state.selected_stream = focused_remote_id_;
+        state.fullscreen = fullscreen_button_pressed_;
+        if (ui_) {
+          state.tab_count = ui_->tab_ids.size();
         }
-
-#if defined(_WIN32)
-        if (state == slint::RenderingState::BeforeRendering &&
-            windows_opengl_video_renderer_ &&
-            windows_opengl_video_renderer_->IsReady() && ui_->stream &&
-            (*ui_->stream)->get_native_video_enabled()) {
-          const auto window_size = (*ui_->stream)->window().size();
-          const float scale_factor =
-              std::max(1.0f, (*ui_->stream)->window().scale_factor());
-          const int top_inset =
-              !fullscreen_button_pressed_ && ui_->tab_ids.size() > 1
-                  ? static_cast<int>(std::lround(30.0f * scale_factor))
-                  : 0;
-          windows_opengl_video_renderer_->RenderLatest(
-              focused_remote_id_, static_cast<int>(window_size.width),
-              static_cast<int>(window_size.height), top_inset);
-          return;
-        }
-#endif
-
-        if (state == slint::RenderingState::BeforeRendering &&
-            ui_->video_gl_texture != 0 && ui_->video_gl_pending_dirty &&
-            !ui_->video_gl_pending_frame.empty()) {
-          GLint previous_texture = 0;
-          GLint previous_unpack_alignment = 0;
-          glGetIntegerv(GL_TEXTURE_BINDING_2D, &previous_texture);
-          glGetIntegerv(GL_UNPACK_ALIGNMENT, &previous_unpack_alignment);
-          glBindTexture(GL_TEXTURE_2D, ui_->video_gl_texture);
-          glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
-          if (ui_->video_gl_texture_size != ui_->video_gl_pending_size) {
-            glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA,
-                         ui_->video_gl_pending_size.width,
-                         ui_->video_gl_pending_size.height, 0, GL_RGBA,
-                         GL_UNSIGNED_BYTE, ui_->video_gl_pending_frame.data());
-            ui_->video_gl_texture_size = ui_->video_gl_pending_size;
-          } else {
-            glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0,
-                            ui_->video_gl_pending_size.width,
-                            ui_->video_gl_pending_size.height, GL_RGBA,
-                            GL_UNSIGNED_BYTE,
-                            ui_->video_gl_pending_frame.data());
-          }
-          glPixelStorei(GL_UNPACK_ALIGNMENT, previous_unpack_alignment);
-          glBindTexture(GL_TEXTURE_2D,
-                        static_cast<GLuint>(previous_texture));
-          ui_->video_gl_pending_dirty = false;
-          return;
-        }
-
-        if (state == slint::RenderingState::RenderingTeardown) {
-#if defined(_WIN32)
-          if (windows_opengl_video_renderer_) {
-            windows_opengl_video_renderer_->Teardown();
-          }
-#endif
-          if (ui_->video_gl_texture != 0) {
-            const GLuint texture = ui_->video_gl_texture;
-            glDeleteTextures(1, &texture);
-          }
-          ui_->video_gl_texture = 0;
-          ui_->video_gl_texture_size = {};
-          ui_->video_gl_image_size = {};
-          ui_->video_gl_pending_size = {};
-          ui_->video_gl_pending_dirty = false;
-          ui_->video_gl_pending_frame.clear();
-        }
-      });
-  if (error.has_value()) {
-    LOG_WARN("Slint OpenGL video renderer unavailable, using pixel buffers");
-#if defined(_WIN32)
-  } else if (windows_opengl_video_renderer_) {
-    windows_opengl_video_renderer_->SetSelectedStream(focused_remote_id_);
-#endif
-  }
-#endif
+        return state;
+      },
+      [this] { video_frame_dirty_.store(true, std::memory_order_release); });
 }
 
-#if defined(_WIN32)
-void GuiApplication::SubmitCachedFrameToOpenGl(
-    const std::shared_ptr<RemoteSession>& session) {
-  if (!session || !windows_opengl_video_renderer_ ||
-      !windows_opengl_video_renderer_->IsReady() ||
-      session->connection_status_.load() != ConnectionStatus::Connected) {
-    return;
-  }
-
-  std::shared_ptr<std::vector<unsigned char>> cached_frame;
-  int cached_width = 0;
-  int cached_height = 0;
-  {
-    std::lock_guard lock(session->video_frame_mutex_);
-    if (session->front_frame_ && !session->front_frame_->empty()) {
-      cached_frame = session->front_frame_;
-      cached_width = session->video_width_;
-      cached_height = session->video_height_;
-    } else if (session->thumbnail_frame_ &&
-               !session->thumbnail_frame_->empty()) {
-      cached_frame = session->thumbnail_frame_;
-      cached_width = session->thumbnail_width_;
-      cached_height = session->thumbnail_height_;
-    }
-  }
-
-  if (cached_frame && cached_width > 0 && cached_height > 0 &&
-      windows_opengl_video_renderer_->SubmitCachedNv12(
-          session->remote_id_, cached_frame->data(), cached_frame->size(),
-          cached_width, cached_height) ==
-          WindowsOpenGlVideoRenderer::SubmitResult::submitted) {
-    video_frame_dirty_.store(true, std::memory_order_release);
-  }
-}
-#endif
-
-#if defined(__APPLE__)
-void GuiApplication::SubmitCachedFrameToMetal(
-    const std::shared_ptr<RemoteSession>& session) {
-  if (!session || !mac_metal_video_renderer_ ||
-      !mac_metal_video_renderer_->IsReady() ||
-      session->connection_status_.load() != ConnectionStatus::Connected) {
-    return;
-  }
-
-  std::shared_ptr<std::vector<unsigned char>> cached_frame;
-  int cached_width = 0;
-  int cached_height = 0;
-  {
-    std::lock_guard lock(session->video_frame_mutex_);
-    if (session->front_frame_ && !session->front_frame_->empty()) {
-      cached_frame = session->front_frame_;
-      cached_width = session->video_width_;
-      cached_height = session->video_height_;
-    } else if (session->thumbnail_frame_ &&
-               !session->thumbnail_frame_->empty()) {
-      cached_frame = session->thumbnail_frame_;
-      cached_width = session->thumbnail_width_;
-      cached_height = session->thumbnail_height_;
-    }
-  }
-
-  if (cached_frame && cached_width > 0 && cached_height > 0 &&
-      mac_metal_video_renderer_->SubmitCachedNv12(
-          session->remote_id_, cached_frame->data(), cached_frame->size(),
-          cached_width, cached_height) ==
-          MacMetalVideoRenderer::SubmitResult::submitted) {
-    video_frame_dirty_.store(true, std::memory_order_release);
-  }
-}
-#endif
-
 void GuiApplication::ScheduleNextVideoFrame() {
   if (!ui_) {
     return;
@@ -3028,28 +2558,19 @@ void GuiApplication::ScheduleNextVideoFrame() {
   if (now >= next_video_frame_time_) {
     const bool frame_dirty =
         video_frame_dirty_.exchange(false, std::memory_order_acq_rel);
-#if defined(__APPLE__)
-    const bool native_render_needed =
-        mac_metal_video_renderer_ && mac_metal_video_renderer_->IsReady() &&
-        mac_metal_video_renderer_->IsAttached() && ui_->stream &&
-        mac_metal_video_renderer_->NeedsSurfaceRedraw();
+    const bool native_render_needed = ui_->stream && ui_->video_presenter &&
+                                      ui_->video_presenter->NeedsRedraw();
     if (frame_dirty || native_render_needed) {
       SyncStreamVideoFrame();
     }
-#else
-    if (frame_dirty) {
-      SyncStreamVideoFrame();
-    }
-#endif
     do {
       next_video_frame_time_ += frame_interval;
     } while (next_video_frame_time_ <= now);
   }
 
   const auto after_render = std::chrono::steady_clock::now();
-  const auto remaining =
-      std::max(next_video_frame_time_ - after_render,
-               std::chrono::steady_clock::duration::zero());
+  const auto remaining = std::max(next_video_frame_time_ - after_render,
+                                  std::chrono::steady_clock::duration::zero());
   const auto delay_ns =
       std::chrono::duration_cast<std::chrono::nanoseconds>(remaining).count();
   const auto delay_ms = std::max<int64_t>(1, (delay_ns + 999'999) / 1'000'000);
@@ -3321,20 +2842,17 @@ void GuiApplication::SelectStreamTab(int index) {
     return;
   }
   const std::string selected_remote_id = ui_->tab_ids[index];
-#if defined(__APPLE__) || defined(_WIN32)
   const bool selection_changed = focused_remote_id_ != selected_remote_id;
   bool renderer_selection_changed = false;
-#endif
   if (!controlled_remote_id_.empty() &&
       controlled_remote_id_ != selected_remote_id) {
     keyboard_.ForceReleasePressedKeys();
   }
   focused_remote_id_ = selected_remote_id;
   controlled_remote_id_ = selected_remote_id;
-#if defined(_WIN32)
-  if (windows_opengl_video_renderer_) {
+  if (ui_->video_presenter) {
     renderer_selection_changed =
-        windows_opengl_video_renderer_->SetSelectedStream(selected_remote_id);
+        ui_->video_presenter->SelectStream(selected_remote_id);
   }
   if (renderer_selection_changed) {
     video_frame_dirty_.store(true, std::memory_order_release);
@@ -3342,18 +2860,6 @@ void GuiApplication::SelectStreamTab(int index) {
   if (selection_changed && ui_->stream) {
     (*ui_->stream)->set_has_frame(false);
   }
-#elif defined(__APPLE__)
-  if (mac_metal_video_renderer_) {
-    renderer_selection_changed =
-        mac_metal_video_renderer_->SetSelectedStream(selected_remote_id);
-  }
-  if (renderer_selection_changed) {
-    video_frame_dirty_.store(true, std::memory_order_release);
-  }
-  if (selection_changed && ui_->stream) {
-    (*ui_->stream)->set_has_frame(false);
-  }
-#endif
   std::shared_ptr<RemoteSession> selected_session;
   {
     std::shared_lock lock(remote_sessions_mutex_);
@@ -3366,20 +2872,11 @@ void GuiApplication::SelectStreamTab(int index) {
     if (selected != remote_sessions_.end()) {
       selected_session = selected->second;
     }
-    start_keyboard_capturer_ =
-        selected_session && selected_session->control_mouse_ &&
-        selected_session->connection_status_.load() ==
-            ConnectionStatus::Connected;
+    start_keyboard_capturer_ = selected_session &&
+                               selected_session->control_mouse_ &&
+                               selected_session->connection_status_.load() ==
+                                   ConnectionStatus::Connected;
   }
-#if defined(_WIN32)
-  if (renderer_selection_changed) {
-    SubmitCachedFrameToOpenGl(selected_session);
-  }
-#elif defined(__APPLE__)
-  if (renderer_selection_changed) {
-    SubmitCachedFrameToMetal(selected_session);
-  }
-#endif
 }
 
 void GuiApplication::ReorderStreamTab(int from, float drop_x, float tab_width) {
@@ -3423,21 +2920,15 @@ void GuiApplication::CloseStreamTab(const std::string& remote_id) {
     std::unique_lock lock(remote_sessions_mutex_);
     remote_sessions_.erase(remote_id);
   }
-  ui_->displayed_frame_sequence.erase(remote_id);
+  if (ui_->video_presenter) {
+    ui_->video_presenter->ForgetStream(remote_id);
+  }
   if (focused_remote_id_ == remote_id) {
     focused_remote_id_.clear();
     controlled_remote_id_.clear();
-#if defined(_WIN32)
-    if (windows_opengl_video_renderer_ &&
-        windows_opengl_video_renderer_->SetSelectedStream({})) {
+    if (ui_->video_presenter && ui_->video_presenter->SelectStream({})) {
       video_frame_dirty_.store(true, std::memory_order_release);
     }
-#elif defined(__APPLE__)
-    if (mac_metal_video_renderer_ &&
-        mac_metal_video_renderer_->SetSelectedStream({})) {
-      video_frame_dirty_.store(true, std::memory_order_release);
-    }
-#endif
   }
 }
 
@@ -3631,20 +3122,14 @@ void GuiApplication::Cleanup() {
   WaitForThumbnailSaveTasks();
   devices_.DestroyAudioOutput();
   if (ui_->stream) {
-#if defined(_WIN32)
-    if (windows_opengl_video_renderer_) {
-      windows_opengl_video_renderer_->SetSelectedStream({});
-    }
-#endif
 #if defined(__APPLE__)
     void* stream_view = (*ui_->stream)->window().appkit_view();
     SetStreamWindowFullscreen(false);
     UnregisterStreamWindow(stream_view);
-    if (mac_metal_video_renderer_) {
-      mac_metal_video_renderer_->SetSelectedStream({});
-      mac_metal_video_renderer_->Detach();
-    }
 #endif
+    if (ui_->video_presenter) {
+      ui_->video_presenter->Detach();
+    }
     (*ui_->stream)->hide();
     ui_->stream.reset();
   }

src/gui/application/gui_application.h

@@ -41,14 +41,6 @@ private:
   void SyncStreamVideoFrame();
   void ScheduleNextVideoFrame();
   void ConfigureStreamVideoRenderer();
-#if defined(_WIN32)
-  void SubmitCachedFrameToOpenGl(
-      const std::shared_ptr<RemoteSession>& session);
-#endif
-#if defined(__APPLE__)
-  void SubmitCachedFrameToMetal(
-      const std::shared_ptr<RemoteSession>& session);
-#endif
   void SyncStreamKeyboardFocus();
   void SetStreamKeyboardFocus(bool focused);
   void SyncServerWindow();

src/gui/platform/metal_video_renderer.h

@@ -8,6 +8,8 @@
 #include <string_view>
 #include <vector>
 
+#include "platform/video_renderer.h"
+
 namespace crossdesk {
 
 // macOS video compositor used by CrossDesk only. MiniRTC continues to expose
@@ -15,51 +17,31 @@ namespace crossdesk {
 // copy into shared Metal buffers and converts NV12 to RGB in the fragment
 // shader. AppKit attachment and rendering are main-thread operations, while
 // SubmitNv12() is safe to call from MiniRTC's decode callback thread.
-class MacMetalVideoRenderer {
+class MacMetalVideoRenderer final : public VideoRenderer {
  public:
-  enum class SubmitResult {
-    submitted,
-    not_selected,
-    dropped,
-    failed,
-  };
-
-  enum class RenderResult {
-    rendered,
-    idle,
-    empty,
-    failed,
-  };
-
-  struct RenderOutcome {
-    RenderResult result = RenderResult::failed;
-    int width = 0;
-    int height = 0;
-    uint64_t sequence = 0;
-  };
-
   MacMetalVideoRenderer();
-  ~MacMetalVideoRenderer();
+  ~MacMetalVideoRenderer() override;
 
   MacMetalVideoRenderer(const MacMetalVideoRenderer&) = delete;
   MacMetalVideoRenderer& operator=(const MacMetalVideoRenderer&) = delete;
 
-  bool IsReady() const;
+  bool IsReady() const override;
+  bool IsActive() const override;
 
   // The selected stream is the only stream uploaded. Frames for background
   // tabs are intentionally dropped to avoid wasting memory bandwidth.
-  bool SetSelectedStream(std::string remote_id);
-  void DiscardStream(std::string_view remote_id);
+  bool SetSelectedStream(std::string remote_id) override;
+  void DiscardStream(std::string_view remote_id) override;
 
   // Copies one tightly packed NV12 frame into a free shared Metal slot.
   SubmitResult SubmitNv12(std::string_view remote_id, const uint8_t* data,
-                          size_t size, int width, int height);
+                          size_t size, int width, int height) override;
 
   // Seeds a newly selected stream from its retained CPU snapshot without
   // replacing a newer decoded frame that is already queued or rendering.
   SubmitResult SubmitCachedNv12(std::string_view remote_id,
                                 const uint8_t* data, size_t size, int width,
-                                int height);
+                                int height) override;
 
   // Attaches a native CAMetalLayer-backed sibling below Slint's NSView and
   // restores the containing AppKit window's ordinary opaque titlebar.
@@ -84,7 +66,7 @@ class MacMetalVideoRenderer {
   // a second per-frame CPU copy during normal playback.
   bool CopyLatestNv12(std::string_view remote_id,
                       std::vector<unsigned char>* output, int* width,
-                      int* height) const;
+                      int* height) const override;
 
  private:
   SubmitResult SubmitNv12Internal(std::string_view remote_id,

src/gui/platform/metal_video_renderer.mm

@@ -602,6 +602,10 @@ bool MacMetalVideoRenderer::IsReady() const {
   return impl_ && impl_->Ready();
 }
 
+bool MacMetalVideoRenderer::IsActive() const {
+  return IsReady() && IsAttached();
+}
+
 bool MacMetalVideoRenderer::SetSelectedStream(std::string remote_id) {
   if (!impl_) {
     return false;

src/gui/platform/opengl_video_renderer.cpp

@@ -1,13 +1,20 @@
-#include "platform/windows_opengl_video_renderer.h"
+#include "platform/opengl_video_renderer.h"
 
-#if !defined(_WIN32)
-#error "WindowsOpenGlVideoRenderer is only available on Windows"
+#if !defined(_WIN32) && !defined(__linux__)
+#error "OpenGlVideoRenderer is only available on Windows and Linux"
 #endif
 
+#if defined(_WIN32)
 #include <windows.h>
+#elif defined(__linux__)
+#include <dlfcn.h>
+#endif
 
 #include <GL/gl.h>
 #include <GL/glext.h>
+#if defined(__linux__)
+#include <GL/glx.h>
+#endif
 
 #include <algorithm>
 #include <array>
@@ -49,6 +56,7 @@ struct SharedFrameState {
 };
 
 template <typename T> bool LoadOpenGlFunction(T *function, const char *name) {
+#if defined(_WIN32)
   PROC address = wglGetProcAddress(name);
   if (address == nullptr || address == reinterpret_cast<PROC>(1) ||
       address == reinterpret_cast<PROC>(2) ||
@@ -58,6 +66,14 @@ template <typename T> bool LoadOpenGlFunction(T *function, const char *name) {
     address = module ? GetProcAddress(module, name) : nullptr;
   }
   *function = reinterpret_cast<T>(address);
+#elif defined(__linux__)
+  void *address = dlsym(RTLD_DEFAULT, name);
+  if (address == nullptr) {
+    address = reinterpret_cast<void *>(
+        glXGetProcAddressARB(reinterpret_cast<const GLubyte *>(name)));
+  }
+  *function = reinterpret_cast<T>(address);
+#endif
   return *function != nullptr;
 }
 
@@ -85,7 +101,10 @@ struct OpenGlFunctions {
   PFNGLGETVERTEXATTRIBPOINTERVPROC get_vertex_attrib_pointer_v = nullptr;
   PFNGLLINKPROGRAMPROC link_program = nullptr;
   PFNGLSHADERSOURCEPROC shader_source = nullptr;
+  PFNGLUNIFORM1FPROC uniform_1f = nullptr;
   PFNGLUNIFORM1IPROC uniform_1i = nullptr;
+  PFNGLUNIFORM2FPROC uniform_2f = nullptr;
+  PFNGLUNIFORM4FPROC uniform_4f = nullptr;
   PFNGLUSEPROGRAMPROC use_program = nullptr;
   PFNGLVERTEXATTRIBPOINTERPROC vertex_attrib_pointer = nullptr;
 
@@ -116,7 +135,10 @@ struct OpenGlFunctions {
                               "glGetVertexAttribPointerv") &&
            LoadOpenGlFunction(&link_program, "glLinkProgram") &&
            LoadOpenGlFunction(&shader_source, "glShaderSource") &&
+           LoadOpenGlFunction(&uniform_1f, "glUniform1f") &&
            LoadOpenGlFunction(&uniform_1i, "glUniform1i") &&
+           LoadOpenGlFunction(&uniform_2f, "glUniform2f") &&
+           LoadOpenGlFunction(&uniform_4f, "glUniform4f") &&
            LoadOpenGlFunction(&use_program, "glUseProgram") &&
            LoadOpenGlFunction(&vertex_attrib_pointer, "glVertexAttribPointer");
   }
@@ -205,7 +227,7 @@ void RestoreVertexAttrib(const OpenGlFunctions &gl, GLuint index,
 
 } // namespace
 
-struct WindowsOpenGlVideoRenderer::Impl {
+struct OpenGlVideoRenderer::Impl {
   OpenGlFunctions gl;
   std::shared_ptr<SharedFrameState> frames =
       std::make_shared<SharedFrameState>();
@@ -218,6 +240,10 @@ struct WindowsOpenGlVideoRenderer::Impl {
   GLint texcoord_location = -1;
   GLint y_texture_location = -1;
   GLint uv_texture_location = -1;
+  GLint target_size_location = -1;
+  GLint video_rect_location = -1;
+  GLint corner_radius_location = -1;
+  GLint video_enabled_location = -1;
   int texture_width = 0;
   int texture_height = 0;
   std::string uploaded_stream;
@@ -232,6 +258,10 @@ struct WindowsOpenGlVideoRenderer::Impl {
     texcoord_location = -1;
     y_texture_location = -1;
     uv_texture_location = -1;
+    target_size_location = -1;
+    video_rect_location = -1;
+    corner_radius_location = -1;
+    video_enabled_location = -1;
     texture_width = 0;
     texture_height = 0;
     uploaded_stream.clear();
@@ -239,12 +269,12 @@ struct WindowsOpenGlVideoRenderer::Impl {
   }
 };
 
-WindowsOpenGlVideoRenderer::WindowsOpenGlVideoRenderer()
+OpenGlVideoRenderer::OpenGlVideoRenderer()
     : impl_(std::make_unique<Impl>()) {}
 
-WindowsOpenGlVideoRenderer::~WindowsOpenGlVideoRenderer() = default;
+OpenGlVideoRenderer::~OpenGlVideoRenderer() = default;
 
-bool WindowsOpenGlVideoRenderer::Setup() {
+bool OpenGlVideoRenderer::Setup() {
   if (IsReady()) {
     return true;
   }
@@ -274,15 +304,41 @@ precision mediump float;
 #endif
 uniform sampler2D y_texture;
 uniform sampler2D uv_texture;
+uniform vec2 target_size;
+uniform vec4 video_rect;
+uniform float corner_radius;
+uniform float video_enabled;
 varying vec2 video_texcoord;
 void main() {
-  float y = 1.16438356 *
-            (texture2D(y_texture, video_texcoord).r - 16.0 / 255.0);
-  vec2 uv = texture2D(uv_texture, video_texcoord).ra - vec2(0.5, 0.5);
-  vec3 rgb = vec3(y + 1.59602678 * uv.y,
-                  y - 0.39176229 * uv.x - 0.81296764 * uv.y,
-                  y + 2.01723214 * uv.x);
-  gl_FragColor = vec4(clamp(rgb, 0.0, 1.0), 1.0);
+  vec2 surface_point = video_texcoord * target_size;
+  vec3 rgb = vec3(0.0);
+  if (video_enabled > 0.5 &&
+      surface_point.x >= video_rect.x &&
+      surface_point.x <= video_rect.x + video_rect.z &&
+      surface_point.y >= video_rect.y &&
+      surface_point.y <= video_rect.y + video_rect.w) {
+    vec2 frame_texcoord =
+        (surface_point - video_rect.xy) / video_rect.zw;
+    float y = 1.16438356 *
+              (texture2D(y_texture, frame_texcoord).r - 16.0 / 255.0);
+    vec2 uv =
+        texture2D(uv_texture, frame_texcoord).ra - vec2(0.5, 0.5);
+    rgb = clamp(vec3(y + 1.59602678 * uv.y,
+                     y - 0.39176229 * uv.x - 0.81296764 * uv.y,
+                     y + 2.01723214 * uv.x),
+                0.0, 1.0);
+  }
+  float coverage = 1.0;
+  if (corner_radius > 0.0) {
+    vec2 half_size = target_size * 0.5;
+    vec2 corner = abs(surface_point - half_size) -
+                  (half_size - vec2(corner_radius));
+    float distance_to_edge =
+        length(max(corner, vec2(0.0))) +
+        min(max(corner.x, corner.y), 0.0) - corner_radius;
+    coverage = clamp(0.5 - distance_to_edge, 0.0, 1.0);
+  }
+  gl_FragColor = vec4(rgb * coverage, coverage);
 }
 )glsl";
   static constexpr char kDesktopVertexShader[] = R"glsl(
@@ -299,15 +355,41 @@ void main() {
 #version 110
 uniform sampler2D y_texture;
 uniform sampler2D uv_texture;
+uniform vec2 target_size;
+uniform vec4 video_rect;
+uniform float corner_radius;
+uniform float video_enabled;
 varying vec2 video_texcoord;
 void main() {
-  float y = 1.16438356 *
-            (texture2D(y_texture, video_texcoord).r - 16.0 / 255.0);
-  vec2 uv = texture2D(uv_texture, video_texcoord).ra - vec2(0.5, 0.5);
-  vec3 rgb = vec3(y + 1.59602678 * uv.y,
-                  y - 0.39176229 * uv.x - 0.81296764 * uv.y,
-                  y + 2.01723214 * uv.x);
-  gl_FragColor = vec4(clamp(rgb, 0.0, 1.0), 1.0);
+  vec2 surface_point = video_texcoord * target_size;
+  vec3 rgb = vec3(0.0);
+  if (video_enabled > 0.5 &&
+      surface_point.x >= video_rect.x &&
+      surface_point.x <= video_rect.x + video_rect.z &&
+      surface_point.y >= video_rect.y &&
+      surface_point.y <= video_rect.y + video_rect.w) {
+    vec2 frame_texcoord =
+        (surface_point - video_rect.xy) / video_rect.zw;
+    float y = 1.16438356 *
+              (texture2D(y_texture, frame_texcoord).r - 16.0 / 255.0);
+    vec2 uv =
+        texture2D(uv_texture, frame_texcoord).ra - vec2(0.5, 0.5);
+    rgb = clamp(vec3(y + 1.59602678 * uv.y,
+                     y - 0.39176229 * uv.x - 0.81296764 * uv.y,
+                     y + 2.01723214 * uv.x),
+                0.0, 1.0);
+  }
+  float coverage = 1.0;
+  if (corner_radius > 0.0) {
+    vec2 half_size = target_size * 0.5;
+    vec2 corner = abs(surface_point - half_size) -
+                  (half_size - vec2(corner_radius));
+    float distance_to_edge =
+        length(max(corner, vec2(0.0))) +
+        min(max(corner.x, corner.y), 0.0) - corner_radius;
+    coverage = clamp(0.5 - distance_to_edge, 0.0, 1.0);
+  }
+  gl_FragColor = vec4(rgb * coverage, coverage);
 }
 )glsl";
 
@@ -363,8 +445,19 @@ void main() {
       impl_->gl.get_uniform_location(impl_->program, "y_texture");
   impl_->uv_texture_location =
       impl_->gl.get_uniform_location(impl_->program, "uv_texture");
+  impl_->target_size_location =
+      impl_->gl.get_uniform_location(impl_->program, "target_size");
+  impl_->video_rect_location =
+      impl_->gl.get_uniform_location(impl_->program, "video_rect");
+  impl_->corner_radius_location =
+      impl_->gl.get_uniform_location(impl_->program, "corner_radius");
+  impl_->video_enabled_location =
+      impl_->gl.get_uniform_location(impl_->program, "video_enabled");
   if (impl_->position_location < 0 || impl_->texcoord_location < 0 ||
-      impl_->y_texture_location < 0 || impl_->uv_texture_location < 0) {
+      impl_->y_texture_location < 0 || impl_->uv_texture_location < 0 ||
+      impl_->target_size_location < 0 || impl_->video_rect_location < 0 ||
+      impl_->corner_radius_location < 0 ||
+      impl_->video_enabled_location < 0) {
     LOG_ERROR("OpenGL NV12 program is missing required shader bindings");
     impl_->gl.delete_program(impl_->program);
     impl_->ResetGlHandles();
@@ -429,7 +522,7 @@ void main() {
   return true;
 }
 
-void WindowsOpenGlVideoRenderer::Teardown() {
+void OpenGlVideoRenderer::Teardown() {
   impl_->ready.store(false, std::memory_order_release);
   if (impl_->y_texture != 0)
     glDeleteTextures(1, &impl_->y_texture);
@@ -444,11 +537,13 @@ void WindowsOpenGlVideoRenderer::Teardown() {
   impl_->ResetGlHandles();
 }
 
-bool WindowsOpenGlVideoRenderer::IsReady() const {
+bool OpenGlVideoRenderer::IsReady() const {
   return impl_->ready.load(std::memory_order_acquire);
 }
 
-bool WindowsOpenGlVideoRenderer::SetSelectedStream(std::string remote_id) {
+bool OpenGlVideoRenderer::IsActive() const { return IsReady(); }
+
+bool OpenGlVideoRenderer::SetSelectedStream(std::string remote_id) {
   std::lock_guard lock(impl_->frames->mutex);
   if (impl_->frames->selected_stream == remote_id) {
     return false;
@@ -463,7 +558,7 @@ bool WindowsOpenGlVideoRenderer::SetSelectedStream(std::string remote_id) {
   return true;
 }
 
-void WindowsOpenGlVideoRenderer::DiscardStream(std::string_view remote_id) {
+void OpenGlVideoRenderer::DiscardStream(std::string_view remote_id) {
   std::lock_guard lock(impl_->frames->mutex);
   for (auto &slot : impl_->frames->slots) {
     if (slot.remote_id == remote_id && slot.use != SlotUse::uploading) {
@@ -477,22 +572,22 @@ void WindowsOpenGlVideoRenderer::DiscardStream(std::string_view remote_id) {
   }
 }
 
-WindowsOpenGlVideoRenderer::SubmitResult
-WindowsOpenGlVideoRenderer::SubmitNv12(std::string_view remote_id,
+OpenGlVideoRenderer::SubmitResult
+OpenGlVideoRenderer::SubmitNv12(std::string_view remote_id,
                                        const uint8_t *data, size_t size,
                                        int width, int height) {
   return SubmitNv12Internal(remote_id, data, size, width, height, true);
 }
 
-WindowsOpenGlVideoRenderer::SubmitResult
-WindowsOpenGlVideoRenderer::SubmitCachedNv12(std::string_view remote_id,
+OpenGlVideoRenderer::SubmitResult
+OpenGlVideoRenderer::SubmitCachedNv12(std::string_view remote_id,
                                              const uint8_t *data, size_t size,
                                              int width, int height) {
   return SubmitNv12Internal(remote_id, data, size, width, height, false);
 }
 
-WindowsOpenGlVideoRenderer::SubmitResult
-WindowsOpenGlVideoRenderer::SubmitNv12Internal(std::string_view remote_id,
+OpenGlVideoRenderer::SubmitResult
+OpenGlVideoRenderer::SubmitNv12Internal(std::string_view remote_id,
                                                const uint8_t *data, size_t size,
                                                int width, int height,
                                                bool replace_pending) {
@@ -562,10 +657,11 @@ WindowsOpenGlVideoRenderer::SubmitNv12Internal(std::string_view remote_id,
   return SubmitResult::submitted;
 }
 
-WindowsOpenGlVideoRenderer::RenderOutcome
-WindowsOpenGlVideoRenderer::RenderLatest(std::string_view remote_id,
-                                         int target_width, int target_height,
-                                         int top_inset_pixels) {
+OpenGlVideoRenderer::RenderOutcome
+OpenGlVideoRenderer::RenderLatest(std::string_view remote_id,
+                                  int target_width, int target_height,
+                                  int top_inset_pixels,
+                                  int corner_radius_pixels) {
   if (!IsReady() || target_width <= 0 || target_height <= 0) {
     return {RenderResult::failed};
   }
@@ -643,7 +739,10 @@ WindowsOpenGlVideoRenderer::RenderLatest(std::string_view remote_id,
   glDisable(GL_STENCIL_TEST);
   glColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
   glViewport(0, 0, target_width, target_height);
-  glClearColor(0.0f, 0.0f, 0.0f, 1.0f);
+  // The Slint items above this underlay are transparent around the custom
+  // window corners. Keep those pixels transparent so the compositor can show
+  // the desktop instead of an opaque black framebuffer clear.
+  glClearColor(0.0f, 0.0f, 0.0f, 0.0f);
   glClear(GL_COLOR_BUFFER_BIT);
 
   bool upload_succeeded = true;
@@ -696,60 +795,77 @@ WindowsOpenGlVideoRenderer::RenderLatest(std::string_view remote_id,
     }
   }
 
-  RenderOutcome outcome{RenderResult::empty};
-  if (upload_succeeded && impl_->uploaded_stream == remote_id &&
-      impl_->texture_width > 0 && impl_->texture_height > 0) {
+  const int content_y = std::clamp(top_inset_pixels, 0, target_height);
+  const int content_height = target_height - content_y;
+  const bool has_video =
+      upload_succeeded && impl_->uploaded_stream == remote_id &&
+      impl_->texture_width > 0 && impl_->texture_height > 0 &&
+      content_height > 0;
+  int video_x = 0;
+  int video_y = content_y;
+  int video_width = target_width;
+  int video_height = content_height;
+  if (has_video) {
     source_width = impl_->texture_width;
     source_height = impl_->texture_height;
     sequence = impl_->uploaded_sequence;
-    const int video_height = std::max(
-        0, target_height - std::clamp(top_inset_pixels, 0, target_height));
-    if (video_height > 0) {
-      const double source_aspect =
-          static_cast<double>(source_width) / source_height;
-      const double target_aspect =
-          static_cast<double>(target_width) / video_height;
-      int viewport_width = target_width;
-      int viewport_height = video_height;
-      int viewport_x = 0;
-      int viewport_y = 0;
-      if (source_aspect > target_aspect) {
-        viewport_height =
-            std::max(1, static_cast<int>(target_width / source_aspect + 0.5));
-        viewport_y = (video_height - viewport_height) / 2;
-      } else {
-        viewport_width =
-            std::max(1, static_cast<int>(video_height * source_aspect + 0.5));
-        viewport_x = (target_width - viewport_width) / 2;
-      }
-
-      glViewport(viewport_x, viewport_y, viewport_width, viewport_height);
-      impl_->gl.use_program(impl_->program);
-      impl_->gl.active_texture(GL_TEXTURE0);
-      glBindTexture(GL_TEXTURE_2D, impl_->y_texture);
-      impl_->gl.uniform_1i(impl_->y_texture_location, 0);
-      impl_->gl.active_texture(GL_TEXTURE1);
-      glBindTexture(GL_TEXTURE_2D, impl_->uv_texture);
-      impl_->gl.uniform_1i(impl_->uv_texture_location, 1);
-      impl_->gl.bind_buffer(GL_ARRAY_BUFFER, impl_->vertex_buffer);
-      impl_->gl.vertex_attrib_pointer(
-          static_cast<GLuint>(impl_->position_location), 2, GL_FLOAT, GL_FALSE,
-          4 * static_cast<GLsizei>(sizeof(GLfloat)), nullptr);
-      impl_->gl.vertex_attrib_pointer(
-          static_cast<GLuint>(impl_->texcoord_location), 2, GL_FLOAT, GL_FALSE,
-          4 * static_cast<GLsizei>(sizeof(GLfloat)),
-          reinterpret_cast<const void *>(2 * sizeof(GLfloat)));
-      impl_->gl.enable_vertex_attrib_array(
-          static_cast<GLuint>(impl_->position_location));
-      impl_->gl.enable_vertex_attrib_array(
-          static_cast<GLuint>(impl_->texcoord_location));
-      glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
-      outcome = {glGetError() == GL_NO_ERROR ? RenderResult::rendered
-                                             : RenderResult::failed,
-                 source_width, source_height, sequence};
+    const double source_aspect =
+        static_cast<double>(source_width) / source_height;
+    const double target_aspect =
+        static_cast<double>(target_width) / content_height;
+    if (source_aspect > target_aspect) {
+      video_height =
+          std::max(1, static_cast<int>(target_width / source_aspect + 0.5));
+      video_y = content_y + (content_height - video_height) / 2;
+    } else {
+      video_width =
+          std::max(1, static_cast<int>(content_height * source_aspect + 0.5));
+      video_x = (target_width - video_width) / 2;
     }
   }
 
+  while (glGetError() != GL_NO_ERROR) {
+  }
+  impl_->gl.use_program(impl_->program);
+  impl_->gl.active_texture(GL_TEXTURE0);
+  glBindTexture(GL_TEXTURE_2D, impl_->y_texture);
+  impl_->gl.uniform_1i(impl_->y_texture_location, 0);
+  impl_->gl.active_texture(GL_TEXTURE1);
+  glBindTexture(GL_TEXTURE_2D, impl_->uv_texture);
+  impl_->gl.uniform_1i(impl_->uv_texture_location, 1);
+  impl_->gl.uniform_2f(impl_->target_size_location,
+                       static_cast<GLfloat>(target_width),
+                       static_cast<GLfloat>(target_height));
+  impl_->gl.uniform_4f(
+      impl_->video_rect_location, static_cast<GLfloat>(video_x),
+      static_cast<GLfloat>(video_y), static_cast<GLfloat>(video_width),
+      static_cast<GLfloat>(video_height));
+  impl_->gl.uniform_1f(
+      impl_->corner_radius_location,
+      static_cast<GLfloat>(std::clamp(
+          corner_radius_pixels, 0, std::min(target_width, target_height) / 2)));
+  impl_->gl.uniform_1f(impl_->video_enabled_location, has_video ? 1.0f : 0.0f);
+  impl_->gl.bind_buffer(GL_ARRAY_BUFFER, impl_->vertex_buffer);
+  impl_->gl.vertex_attrib_pointer(
+      static_cast<GLuint>(impl_->position_location), 2, GL_FLOAT, GL_FALSE,
+      4 * static_cast<GLsizei>(sizeof(GLfloat)), nullptr);
+  impl_->gl.vertex_attrib_pointer(
+      static_cast<GLuint>(impl_->texcoord_location), 2, GL_FLOAT, GL_FALSE,
+      4 * static_cast<GLsizei>(sizeof(GLfloat)),
+      reinterpret_cast<const void *>(2 * sizeof(GLfloat)));
+  impl_->gl.enable_vertex_attrib_array(
+      static_cast<GLuint>(impl_->position_location));
+  impl_->gl.enable_vertex_attrib_array(
+      static_cast<GLuint>(impl_->texcoord_location));
+  glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
+  const GLenum draw_error = glGetError();
+  RenderOutcome outcome{
+      !upload_succeeded || draw_error != GL_NO_ERROR
+          ? RenderResult::failed
+          : has_video ? RenderResult::rendered : RenderResult::empty,
+      has_video ? source_width : 0, has_video ? source_height : 0,
+      has_video ? sequence : 0};
+
   RestoreVertexAttrib(impl_->gl, static_cast<GLuint>(impl_->position_location),
                       position_state);
   RestoreVertexAttrib(impl_->gl, static_cast<GLuint>(impl_->texcoord_location),
@@ -786,7 +902,7 @@ WindowsOpenGlVideoRenderer::RenderLatest(std::string_view remote_id,
   return outcome;
 }
 
-bool WindowsOpenGlVideoRenderer::CopyLatestNv12(
+bool OpenGlVideoRenderer::CopyLatestNv12(
     std::string_view remote_id, std::vector<unsigned char> *output, int *width,
     int *height) const {
   if (!output || !width || !height) {

src/gui/platform/opengl_video_renderer.h

@@ -0,0 +1,70 @@
+#ifndef CROSSDESK_GUI_PLATFORM_OPENGL_VIDEO_RENDERER_H_
+#define CROSSDESK_GUI_PLATFORM_OPENGL_VIDEO_RENDERER_H_
+
+#include <cstddef>
+#include <cstdint>
+#include <memory>
+#include <string>
+#include <string_view>
+#include <vector>
+
+#include "platform/video_renderer.h"
+
+namespace crossdesk {
+
+// NV12 underlay for Slint's FemtoVG OpenGL renderer on Windows and Linux.
+// MiniRTC submits tightly packed CPU NV12 frames from its decode callback
+// thread. The Slint UI thread uploads the newest frame as Y and UV textures and
+// performs color conversion and scaling in a fragment shader before Slint
+// draws its overlay.
+class OpenGlVideoRenderer final : public VideoRenderer {
+public:
+  OpenGlVideoRenderer();
+  ~OpenGlVideoRenderer() override;
+
+  OpenGlVideoRenderer(const OpenGlVideoRenderer &) = delete;
+  OpenGlVideoRenderer &operator=(const OpenGlVideoRenderer &) = delete;
+
+  // These methods must run while Slint's OpenGL context is current, from the
+  // RenderingSetup and RenderingTeardown notifier states respectively.
+  bool Setup();
+  void Teardown();
+  bool IsReady() const override;
+  bool IsActive() const override;
+
+  // Only the selected stream is queued for upload. Frames for other tabs are
+  // retained by GuiRuntime at a throttled rate for thumbnail/tab restoration.
+  bool SetSelectedStream(std::string remote_id) override;
+  void DiscardStream(std::string_view remote_id) override;
+
+  // Thread-safe; called from MiniRTC's decode callback thread.
+  SubmitResult SubmitNv12(std::string_view remote_id, const uint8_t *data,
+                          size_t size, int width, int height) override;
+  SubmitResult SubmitCachedNv12(std::string_view remote_id, const uint8_t *data,
+                                size_t size, int width, int height) override;
+
+  // Draws below Slint while its OpenGL context is current. All dimensions are
+  // physical pixels in the window client area. corner_radius_pixels clips the
+  // underlay to Slint's custom rounded window surface.
+  RenderOutcome RenderLatest(std::string_view remote_id, int target_width,
+                             int target_height, int top_inset_pixels,
+                             int corner_radius_pixels);
+
+  // Used during disconnect cleanup so thumbnail generation does not require a
+  // second per-frame CPU copy during normal playback.
+  bool CopyLatestNv12(std::string_view remote_id,
+                      std::vector<unsigned char> *output, int *width,
+                      int *height) const override;
+
+private:
+  SubmitResult SubmitNv12Internal(std::string_view remote_id,
+                                  const uint8_t *data, size_t size, int width,
+                                  int height, bool replace_pending);
+
+  struct Impl;
+  std::unique_ptr<Impl> impl_;
+};
+
+} // namespace crossdesk
+
+#endif // CROSSDESK_GUI_PLATFORM_OPENGL_VIDEO_RENDERER_H_

src/gui/platform/video_renderer.h

@@ -0,0 +1,70 @@
+#ifndef CROSSDESK_GUI_PLATFORM_VIDEO_RENDERER_H_
+#define CROSSDESK_GUI_PLATFORM_VIDEO_RENDERER_H_
+
+#include <cstddef>
+#include <cstdint>
+#include <memory>
+#include <string>
+#include <string_view>
+#include <vector>
+
+namespace crossdesk {
+
+// Platform-independent contract for the native stream video renderer.
+//
+// Frame submission and snapshot methods may be called from MiniRTC callback
+// threads. Surface setup, rendering and teardown remain backend-specific and
+// are driven by the Slint presentation module on the UI thread.
+class VideoRenderer {
+public:
+  enum class SubmitResult {
+    submitted,
+    not_selected,
+    dropped,
+    failed,
+  };
+
+  enum class RenderResult {
+    rendered,
+    idle,
+    empty,
+    failed,
+  };
+
+  struct RenderOutcome {
+    RenderResult result = RenderResult::failed;
+    int width = 0;
+    int height = 0;
+    uint64_t sequence = 0;
+  };
+
+  virtual ~VideoRenderer() = default;
+
+  virtual bool IsReady() const = 0;
+
+  // True only while the renderer can accept and present native frames. Metal
+  // additionally requires attachment to an AppKit view; OpenGL requires an
+  // active Slint rendering context.
+  virtual bool IsActive() const = 0;
+
+  virtual bool SetSelectedStream(std::string remote_id) = 0;
+  virtual void DiscardStream(std::string_view remote_id) = 0;
+
+  virtual SubmitResult SubmitNv12(std::string_view remote_id,
+                                  const uint8_t* data, size_t size, int width,
+                                  int height) = 0;
+  virtual SubmitResult SubmitCachedNv12(std::string_view remote_id,
+                                        const uint8_t* data, size_t size,
+                                        int width, int height) = 0;
+
+  virtual bool CopyLatestNv12(std::string_view remote_id,
+                              std::vector<unsigned char>* output, int* width,
+                              int* height) const = 0;
+};
+
+// Exactly one factory implementation is linked for the target platform.
+std::unique_ptr<VideoRenderer> CreateVideoRenderer();
+
+}  // namespace crossdesk
+
+#endif  // CROSSDESK_GUI_PLATFORM_VIDEO_RENDERER_H_

src/gui/platform/video_renderer_factory_metal.mm

@@ -0,0 +1,13 @@
+#include "platform/video_renderer.h"
+
+#include <memory>
+
+#include "platform/metal_video_renderer.h"
+
+namespace crossdesk {
+
+std::unique_ptr<VideoRenderer> CreateVideoRenderer() {
+  return std::make_unique<MacMetalVideoRenderer>();
+}
+
+}  // namespace crossdesk

src/gui/platform/video_renderer_factory_opengl.cpp

@@ -0,0 +1,12 @@
+#include <memory>
+
+#include "platform/opengl_video_renderer.h"
+#include "platform/video_renderer.h"
+
+namespace crossdesk {
+
+std::unique_ptr<VideoRenderer> CreateVideoRenderer() {
+  return std::make_unique<OpenGlVideoRenderer>();
+}
+
+}  // namespace crossdesk

src/gui/platform/windows_opengl_video_renderer.h

@@ -1,86 +0,0 @@
-#ifndef CROSSDESK_GUI_PLATFORM_WINDOWS_OPENGL_VIDEO_RENDERER_H_
-#define CROSSDESK_GUI_PLATFORM_WINDOWS_OPENGL_VIDEO_RENDERER_H_
-
-#include <cstddef>
-#include <cstdint>
-#include <memory>
-#include <string>
-#include <string_view>
-#include <vector>
-
-namespace crossdesk {
-
-// Windows NV12 underlay for Slint's FemtoVG OpenGL renderer. MiniRTC submits
-// tightly packed CPU NV12 frames from its decode callback thread. The Slint UI
-// thread uploads the newest frame as Y and UV textures and performs color
-// conversion and scaling in a fragment shader before Slint draws its overlay.
-class WindowsOpenGlVideoRenderer {
-public:
-  enum class SubmitResult {
-    submitted,
-    not_selected,
-    dropped,
-    failed,
-  };
-
-  enum class RenderResult {
-    rendered,
-    idle,
-    empty,
-    failed,
-  };
-
-  struct RenderOutcome {
-    RenderResult result = RenderResult::failed;
-    int width = 0;
-    int height = 0;
-    uint64_t sequence = 0;
-  };
-
-  WindowsOpenGlVideoRenderer();
-  ~WindowsOpenGlVideoRenderer();
-
-  WindowsOpenGlVideoRenderer(const WindowsOpenGlVideoRenderer &) = delete;
-  WindowsOpenGlVideoRenderer &
-  operator=(const WindowsOpenGlVideoRenderer &) = delete;
-
-  // These methods must run while Slint's OpenGL context is current, from the
-  // RenderingSetup and RenderingTeardown notifier states respectively.
-  bool Setup();
-  void Teardown();
-  bool IsReady() const;
-
-  // Only the selected stream is queued for upload. Frames for other tabs are
-  // retained by GuiRuntime at a throttled rate for thumbnail/tab restoration.
-  bool SetSelectedStream(std::string remote_id);
-  void DiscardStream(std::string_view remote_id);
-
-  // Thread-safe; called from MiniRTC's decode callback thread.
-  SubmitResult SubmitNv12(std::string_view remote_id, const uint8_t *data,
-                          size_t size, int width, int height);
-  SubmitResult SubmitCachedNv12(std::string_view remote_id, const uint8_t *data,
-                                size_t size, int width, int height);
-
-  // Draws below Slint while its OpenGL context is current. target dimensions
-  // and top_inset_pixels are physical pixels in the window client area.
-  RenderOutcome RenderLatest(std::string_view remote_id, int target_width,
-                             int target_height, int top_inset_pixels);
-
-  // Used during disconnect cleanup so thumbnail generation does not require a
-  // second per-frame CPU copy during normal playback.
-  bool CopyLatestNv12(std::string_view remote_id,
-                      std::vector<unsigned char> *output, int *width,
-                      int *height) const;
-
-private:
-  SubmitResult SubmitNv12Internal(std::string_view remote_id,
-                                  const uint8_t *data, size_t size, int width,
-                                  int height, bool replace_pending);
-
-  struct Impl;
-  std::unique_ptr<Impl> impl_;
-};
-
-} // namespace crossdesk
-
-#endif // CROSSDESK_GUI_PLATFORM_WINDOWS_OPENGL_VIDEO_RENDERER_H_

src/gui/rendering/slint_video_presenter.cpp

@@ -0,0 +1,502 @@
+#include "rendering/slint_video_presenter.h"
+
+#include <libyuv.h>
+#include <slint.h>
+
+#include <algorithm>
+#include <cmath>
+#include <mutex>
+#include <unordered_map>
+#include <utility>
+
+#include "nv12_scaler.h"
+#include "platform/video_renderer.h"
+#include "rd_log.h"
+
+#if defined(__APPLE__)
+#include "platform/metal_video_renderer.h"
+#else
+#if defined(_WIN32)
+#include <windows.h>
+#endif
+#include <GL/gl.h>
+
+#include "platform/opengl_video_renderer.h"
+#endif
+
+namespace crossdesk {
+namespace {
+
+#if !defined(__APPLE__)
+constexpr GLint kGlClampToEdge = 0x812F;
+constexpr float kStreamWindowCornerRadius = 12.0f;
+#else
+constexpr int kMetalAttachmentAttemptLimit = 30;
+#endif
+
+struct VideoRenderSize {
+  int width = 0;
+  int height = 0;
+
+  bool operator==(const VideoRenderSize&) const = default;
+};
+
+VideoRenderSize FitVideoToRenderArea(int source_width, int source_height,
+                                     int area_width, int area_height) {
+  if (source_width <= 0 || source_height <= 0 || area_width <= 0 ||
+      area_height <= 0) {
+    return {};
+  }
+
+  const double scale =
+      std::min({1.0, static_cast<double>(area_width) / source_width,
+                static_cast<double>(area_height) / source_height});
+  int width = std::max(2, static_cast<int>(std::floor(source_width * scale)));
+  int height = std::max(2, static_cast<int>(std::floor(source_height * scale)));
+
+  width &= ~1;
+  height &= ~1;
+  return {std::min(width, source_width & ~1),
+          std::min(height, source_height & ~1)};
+}
+
+}  // namespace
+
+struct SlintVideoPresenter::Impl {
+  explicit Impl(VideoRenderer& renderer) : renderer(renderer) {}
+
+  VideoRenderer& renderer;
+  slint::ComponentHandle<ui::StreamWindow>* stream = nullptr;
+  StateProvider state_provider;
+  DirtyCallback dirty_callback;
+  std::unordered_map<std::string, uint64_t> displayed_frame_sequence;
+  std::unordered_map<std::string, uint64_t> seeded_frame_sequence;
+  std::vector<uint8_t> scaled_video_frame;
+  std::vector<uint8_t> scaled_video_frame_scratch;
+
+#if defined(__APPLE__)
+  int attachment_attempts = 0;
+  bool skip_attachment_once = false;
+#else
+  std::mutex gl_mutex;
+  std::vector<uint8_t> gl_conversion_frame;
+  std::vector<uint8_t> gl_pending_frame;
+  uint32_t gl_texture = 0;
+  VideoRenderSize gl_texture_size;
+  VideoRenderSize gl_image_size;
+  VideoRenderSize gl_pending_size;
+  bool gl_pending_dirty = false;
+#endif
+
+  SurfaceState CurrentState() const {
+    return state_provider ? state_provider() : SurfaceState{};
+  }
+
+  void MarkDirty() const {
+    if (dirty_callback) {
+      dirty_callback();
+    }
+  }
+
+  bool MarkPresented(const std::string& remote_id, uint64_t sequence) {
+    if (remote_id.empty() || sequence == 0 ||
+        displayed_frame_sequence[remote_id] == sequence) {
+      return false;
+    }
+    displayed_frame_sequence[remote_id] = sequence;
+    return true;
+  }
+
+#if !defined(__APPLE__)
+  void ConfigureOpenGlNotifier() {
+    if (!stream) {
+      return;
+    }
+
+    const auto error = (*stream)->window().set_rendering_notifier(
+        [this](slint::RenderingState state, slint::GraphicsAPI graphics_api) {
+          if (graphics_api != slint::GraphicsAPI::NativeOpenGL) {
+            return;
+          }
+
+          std::lock_guard lock(gl_mutex);
+          auto* opengl_renderer =
+              dynamic_cast<OpenGlVideoRenderer*>(&renderer);
+          if (state == slint::RenderingState::RenderingSetup) {
+            GLuint texture = 0;
+            glGenTextures(1, &texture);
+            if (texture == 0) {
+              return;
+            }
+            GLint previous_texture = 0;
+            glGetIntegerv(GL_TEXTURE_BINDING_2D, &previous_texture);
+            glBindTexture(GL_TEXTURE_2D, texture);
+            glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
+            glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
+            glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, kGlClampToEdge);
+            glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, kGlClampToEdge);
+            glBindTexture(GL_TEXTURE_2D, static_cast<GLuint>(previous_texture));
+            gl_texture = texture;
+            if (opengl_renderer && opengl_renderer->Setup()) {
+              const std::string selected_stream =
+                  CurrentState().selected_stream;
+              opengl_renderer->SetSelectedStream(selected_stream);
+              seeded_frame_sequence.erase(selected_stream);
+              MarkDirty();
+            }
+            return;
+          }
+
+          if (state == slint::RenderingState::BeforeRendering && stream &&
+              opengl_renderer && opengl_renderer->IsReady() &&
+              (*stream)->get_native_video_enabled()) {
+            const SurfaceState surface = CurrentState();
+            const auto window_size = (*stream)->window().size();
+            const float scale_factor =
+                std::max(1.0f, (*stream)->window().scale_factor());
+            const int top_inset =
+                !surface.fullscreen && surface.tab_count > 1
+                    ? static_cast<int>(std::lround(30.0f * scale_factor))
+                    : 0;
+            const bool rounded_window = (*stream)->get_custom_titlebar() &&
+                                        !surface.fullscreen &&
+                                        !(*stream)->window().is_maximized();
+            const int corner_radius =
+                rounded_window ? static_cast<int>(std::lround(
+                                     kStreamWindowCornerRadius * scale_factor))
+                               : 0;
+            opengl_renderer->RenderLatest(
+                surface.selected_stream, static_cast<int>(window_size.width),
+                static_cast<int>(window_size.height), top_inset, corner_radius);
+            return;
+          }
+
+          if (state == slint::RenderingState::BeforeRendering &&
+              gl_texture != 0 && gl_pending_dirty &&
+              !gl_pending_frame.empty()) {
+            GLint previous_texture = 0;
+            GLint previous_unpack_alignment = 0;
+            glGetIntegerv(GL_TEXTURE_BINDING_2D, &previous_texture);
+            glGetIntegerv(GL_UNPACK_ALIGNMENT, &previous_unpack_alignment);
+            glBindTexture(GL_TEXTURE_2D, gl_texture);
+            glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
+            if (gl_texture_size != gl_pending_size) {
+              glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, gl_pending_size.width,
+                           gl_pending_size.height, 0, GL_RGBA, GL_UNSIGNED_BYTE,
+                           gl_pending_frame.data());
+              gl_texture_size = gl_pending_size;
+            } else {
+              glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, gl_pending_size.width,
+                              gl_pending_size.height, GL_RGBA, GL_UNSIGNED_BYTE,
+                              gl_pending_frame.data());
+            }
+            glPixelStorei(GL_UNPACK_ALIGNMENT, previous_unpack_alignment);
+            glBindTexture(GL_TEXTURE_2D, static_cast<GLuint>(previous_texture));
+            gl_pending_dirty = false;
+            return;
+          }
+
+          if (state == slint::RenderingState::RenderingTeardown) {
+            if (opengl_renderer) {
+              opengl_renderer->Teardown();
+            }
+            if (gl_texture != 0) {
+              const GLuint texture = gl_texture;
+              glDeleteTextures(1, &texture);
+            }
+            gl_texture = 0;
+            gl_texture_size = {};
+            gl_image_size = {};
+            gl_pending_size = {};
+            gl_pending_dirty = false;
+            gl_pending_frame.clear();
+          }
+        });
+
+    if (error.has_value()) {
+      LOG_WARN("Slint OpenGL video renderer unavailable, using pixel buffers");
+    } else {
+      renderer.SetSelectedStream(CurrentState().selected_stream);
+    }
+  }
+#endif
+};
+
+SlintVideoPresenter::SlintVideoPresenter(VideoRenderer& renderer)
+    : impl_(std::make_unique<Impl>(renderer)) {}
+
+SlintVideoPresenter::~SlintVideoPresenter() = default;
+
+void SlintVideoPresenter::PrepareWindow(
+    slint::ComponentHandle<ui::StreamWindow>& stream,
+    StateProvider state_provider, DirtyCallback dirty_callback) {
+  impl_->stream = &stream;
+  impl_->state_provider = std::move(state_provider);
+  impl_->dirty_callback = std::move(dirty_callback);
+#if defined(__APPLE__)
+  stream->set_native_video_enabled(false);
+  impl_->attachment_attempts = kMetalAttachmentAttemptLimit;
+  impl_->skip_attachment_once = true;
+#else
+  impl_->ConfigureOpenGlNotifier();
+#endif
+}
+
+bool SlintVideoPresenter::EnsureAttached() {
+#if defined(__APPLE__)
+  if (!impl_->stream || impl_->attachment_attempts <= 0) {
+    return false;
+  }
+  if (impl_->skip_attachment_once) {
+    impl_->skip_attachment_once = false;
+    return false;
+  }
+
+  auto* metal_renderer =
+      dynamic_cast<MacMetalVideoRenderer*>(&impl_->renderer);
+  if (!metal_renderer) {
+    impl_->attachment_attempts = 0;
+    (*impl_->stream)->set_native_video_enabled(false);
+    return false;
+  }
+  if (!metal_renderer->IsReady()) {
+    impl_->attachment_attempts = 0;
+    (*impl_->stream)->set_native_video_enabled(false);
+    return false;
+  }
+  if (metal_renderer->IsAttached()) {
+    impl_->attachment_attempts = 0;
+    return false;
+  }
+
+  if (void* view = (*impl_->stream)->window().appkit_view()) {
+    (*impl_->stream)->set_native_video_enabled(true);
+    if (metal_renderer->Attach(view)) {
+      impl_->attachment_attempts = 0;
+      const SurfaceState surface = impl_->CurrentState();
+      impl_->displayed_frame_sequence.erase(surface.selected_stream);
+      impl_->seeded_frame_sequence.erase(surface.selected_stream);
+      if (metal_renderer->SetSelectedStream(surface.selected_stream)) {
+        impl_->MarkDirty();
+      }
+      return true;
+    }
+    (*impl_->stream)->set_native_video_enabled(false);
+  }
+
+  if (--impl_->attachment_attempts == 0) {
+    (*impl_->stream)->set_native_video_enabled(false);
+    LOG_WARN("Unable to attach the Metal video surface; using CPU rendering");
+  }
+#endif
+  return false;
+}
+
+bool SlintVideoPresenter::NeedsRedraw() const {
+#if defined(__APPLE__)
+  const auto* metal_renderer =
+      dynamic_cast<const MacMetalVideoRenderer*>(&impl_->renderer);
+  return metal_renderer && metal_renderer->IsActive() &&
+         metal_renderer->NeedsSurfaceRedraw();
+#else
+  return false;
+#endif
+}
+
+bool SlintVideoPresenter::SelectStream(std::string remote_id) {
+  const bool changed = impl_->renderer.SetSelectedStream(remote_id);
+  if (changed && !remote_id.empty()) {
+    impl_->seeded_frame_sequence.erase(remote_id);
+  }
+  return changed;
+}
+
+SlintVideoPresenter::PresentResult SlintVideoPresenter::Present(
+    const Frame& frame) {
+  PresentResult result;
+  if (!impl_->stream || frame.remote_id.empty()) {
+    return result;
+  }
+
+  if (impl_->renderer.IsActive() && frame.nv12 && frame.width > 0 &&
+      frame.height > 0 && frame.sequence > 0 &&
+      impl_->seeded_frame_sequence[frame.remote_id] != frame.sequence) {
+    const auto submit_result = impl_->renderer.SubmitCachedNv12(
+        frame.remote_id, frame.nv12->data(), frame.nv12->size(), frame.width,
+        frame.height);
+    if (submit_result == VideoRenderer::SubmitResult::submitted ||
+        submit_result == VideoRenderer::SubmitResult::dropped) {
+      impl_->seeded_frame_sequence[frame.remote_id] = frame.sequence;
+    }
+  }
+
+#if defined(__APPLE__)
+  auto* metal_renderer =
+      dynamic_cast<MacMetalVideoRenderer*>(&impl_->renderer);
+  if (metal_renderer && metal_renderer->IsActive()) {
+    const SurfaceState surface = impl_->CurrentState();
+    const double top_inset =
+        !surface.fullscreen && surface.tab_count > 1 ? 30.0 : 0.0;
+    const auto outcome = metal_renderer->RenderLatest(
+        frame.remote_id, top_inset, !surface.fullscreen);
+    if (outcome.result == VideoRenderer::RenderResult::rendered) {
+      (*impl_->stream)->set_has_frame(true);
+      (*impl_->stream)->set_receiving_text("");
+      result.width = outcome.width;
+      result.height = outcome.height;
+      result.new_frame =
+          impl_->MarkPresented(frame.remote_id, outcome.sequence);
+    } else if (outcome.result == VideoRenderer::RenderResult::empty) {
+      (*impl_->stream)->set_has_frame(false);
+    }
+    return result;
+  }
+#else
+  auto* opengl_renderer =
+      dynamic_cast<OpenGlVideoRenderer*>(&impl_->renderer);
+  if (opengl_renderer && opengl_renderer->IsReady()) {
+    opengl_renderer->SetSelectedStream(frame.remote_id);
+    if (!(*impl_->stream)->get_native_video_enabled()) {
+      (*impl_->stream)->set_native_video_enabled(true);
+    }
+
+    const bool has_frame =
+        frame.width > 0 && frame.height > 0 && frame.sequence > 0;
+    (*impl_->stream)->set_has_frame(has_frame);
+    if (has_frame) {
+      (*impl_->stream)->set_receiving_text("");
+      result.width = frame.width;
+      result.height = frame.height;
+      result.new_frame = impl_->MarkPresented(frame.remote_id, frame.sequence);
+    }
+    (*impl_->stream)->window().request_redraw();
+    return result;
+  }
+  if ((*impl_->stream)->get_native_video_enabled()) {
+    (*impl_->stream)->set_native_video_enabled(false);
+  }
+#endif
+
+  const size_t nv12_size =
+      static_cast<size_t>(frame.width) * frame.height * 3 / 2;
+  if (!frame.nv12 || frame.width <= 0 || frame.height <= 0 ||
+      frame.nv12->size() < nv12_size) {
+    (*impl_->stream)->set_has_frame(false);
+    return result;
+  }
+  if (impl_->displayed_frame_sequence[frame.remote_id] == frame.sequence) {
+    return result;
+  }
+
+  const auto window_size = (*impl_->stream)->window().size();
+  const VideoRenderSize render_size = FitVideoToRenderArea(
+      frame.width, frame.height, static_cast<int>(window_size.width),
+      static_cast<int>(window_size.height));
+  const uint8_t* y_plane = frame.nv12->data();
+  const uint8_t* uv_plane =
+      frame.nv12->data() + static_cast<size_t>(frame.width) * frame.height;
+  int output_width = frame.width;
+  int output_height = frame.height;
+
+  if (render_size.width > 0 && render_size.height > 0 &&
+      (render_size.width != frame.width ||
+       render_size.height != frame.height)) {
+    const size_t scaled_nv12_size =
+        static_cast<size_t>(render_size.width) * render_size.height * 3 / 2;
+    impl_->scaled_video_frame.resize(scaled_nv12_size);
+    uint8_t* scaled_y = impl_->scaled_video_frame.data();
+    uint8_t* scaled_uv =
+        scaled_y + static_cast<size_t>(render_size.width) * render_size.height;
+    if (ScaleNv12ViaI420(y_plane, frame.width, uv_plane, frame.width,
+                         frame.width, frame.height, scaled_y, render_size.width,
+                         scaled_uv, render_size.width, render_size.width,
+                         render_size.height, libyuv::kFilterBox,
+                         &impl_->scaled_video_frame_scratch) == 0) {
+      y_plane = scaled_y;
+      uv_plane = scaled_uv;
+      output_width = render_size.width;
+      output_height = render_size.height;
+    }
+  }
+
+#if !defined(__APPLE__)
+  uint32_t texture_id = 0;
+  {
+    std::lock_guard lock(impl_->gl_mutex);
+    texture_id = impl_->gl_texture;
+  }
+  if (texture_id != 0) {
+    impl_->gl_conversion_frame.resize(static_cast<size_t>(output_width) *
+                                      output_height * 4);
+    if (libyuv::NV12ToABGR(y_plane, output_width, uv_plane, output_width,
+                           impl_->gl_conversion_frame.data(), output_width * 4,
+                           output_width, output_height) != 0) {
+      return result;
+    }
+    {
+      std::lock_guard lock(impl_->gl_mutex);
+      if (impl_->gl_texture == 0) {
+        return result;
+      }
+      texture_id = impl_->gl_texture;
+      impl_->gl_pending_frame.swap(impl_->gl_conversion_frame);
+      impl_->gl_pending_size = {output_width, output_height};
+      impl_->gl_pending_dirty = true;
+    }
+    const VideoRenderSize output_size{output_width, output_height};
+    if (impl_->gl_image_size != output_size) {
+      (*impl_->stream)
+          ->set_frame(slint::Image::create_from_borrowed_gl_2d_rgba_texture(
+              texture_id,
+              slint::Size<uint32_t>{static_cast<uint32_t>(output_width),
+                                    static_cast<uint32_t>(output_height)}));
+      impl_->gl_image_size = output_size;
+    }
+    (*impl_->stream)->set_has_frame(true);
+    (*impl_->stream)->set_receiving_text("");
+    (*impl_->stream)->window().request_redraw();
+    result.width = frame.width;
+    result.height = frame.height;
+    result.new_frame = impl_->MarkPresented(frame.remote_id, frame.sequence);
+    return result;
+  }
+#endif
+
+  slint::SharedPixelBuffer<slint::Rgb8Pixel> pixels(output_width,
+                                                    output_height);
+  if (libyuv::NV12ToRAW(y_plane, output_width, uv_plane, output_width,
+                        reinterpret_cast<uint8_t*>(pixels.begin()),
+                        output_width * 3, output_width, output_height) != 0) {
+    return result;
+  }
+
+  (*impl_->stream)->set_frame(slint::Image(std::move(pixels)));
+  (*impl_->stream)->set_has_frame(true);
+  (*impl_->stream)->set_receiving_text("");
+  result.width = frame.width;
+  result.height = frame.height;
+  result.new_frame = impl_->MarkPresented(frame.remote_id, frame.sequence);
+  return result;
+}
+
+void SlintVideoPresenter::ForgetStream(const std::string& remote_id) {
+  impl_->displayed_frame_sequence.erase(remote_id);
+  impl_->seeded_frame_sequence.erase(remote_id);
+}
+
+void SlintVideoPresenter::Detach() {
+  impl_->renderer.SetSelectedStream({});
+#if defined(__APPLE__)
+  if (auto* metal_renderer =
+          dynamic_cast<MacMetalVideoRenderer*>(&impl_->renderer)) {
+    metal_renderer->Detach();
+  }
+  impl_->attachment_attempts = 0;
+  impl_->skip_attachment_once = false;
+#endif
+  impl_->stream = nullptr;
+  impl_->state_provider = {};
+  impl_->dirty_callback = {};
+}
+
+}  // namespace crossdesk

src/gui/rendering/slint_video_presenter.h

@@ -0,0 +1,75 @@
+#ifndef CROSSDESK_GUI_RENDERING_SLINT_VIDEO_PRESENTER_H_
+#define CROSSDESK_GUI_RENDERING_SLINT_VIDEO_PRESENTER_H_
+
+#include <cstddef>
+#include <cstdint>
+#include <functional>
+#include <memory>
+#include <string>
+#include <vector>
+
+#include "crossdesk_ui.h"
+
+namespace crossdesk {
+
+class VideoRenderer;
+
+// Owns the bridge between decoded NV12 frames, the selected native graphics
+// backend and Slint's StreamWindow. Platform macros and graphics-API lifecycle
+// details are intentionally contained in the implementation file.
+class SlintVideoPresenter {
+public:
+  struct SurfaceState {
+    std::string selected_stream;
+    bool fullscreen = false;
+    size_t tab_count = 0;
+  };
+
+  struct Frame {
+    std::string remote_id;
+    std::shared_ptr<std::vector<unsigned char>> nv12;
+    int width = 0;
+    int height = 0;
+    uint64_t sequence = 0;
+  };
+
+  struct PresentResult {
+    int width = 0;
+    int height = 0;
+    bool new_frame = false;
+  };
+
+  using StateProvider = std::function<SurfaceState()>;
+  using DirtyCallback = std::function<void()>;
+
+  explicit SlintVideoPresenter(VideoRenderer& renderer);
+  ~SlintVideoPresenter();
+
+  SlintVideoPresenter(const SlintVideoPresenter&) = delete;
+  SlintVideoPresenter& operator=(const SlintVideoPresenter&) = delete;
+
+  // Called before the window is shown. OpenGL installs Slint's rendering
+  // notifier here; macOS deliberately defers native view attachment until a
+  // later UI pass, after AppKit has committed the ordinary window style.
+  void PrepareWindow(slint::ComponentHandle<ui::StreamWindow>& stream,
+                     StateProvider state_provider,
+                     DirtyCallback dirty_callback);
+
+  // Returns true only when a deferred native surface became attached during
+  // this call, allowing the caller to seed it from a cached frame once.
+  bool EnsureAttached();
+  bool NeedsRedraw() const;
+
+  bool SelectStream(std::string remote_id);
+  PresentResult Present(const Frame& frame);
+  void ForgetStream(const std::string& remote_id);
+  void Detach();
+
+private:
+  struct Impl;
+  std::unique_ptr<Impl> impl_;
+};
+
+}  // namespace crossdesk
+
+#endif  // CROSSDESK_GUI_RENDERING_SLINT_VIDEO_PRESENTER_H_

src/gui/runtime/connection_runtime.cpp

@@ -11,12 +11,7 @@
 #include "display_stream_id.h"
 #include "localization.h"
 #include "platform.h"
-#if defined(_WIN32)
-#include "platform/windows_opengl_video_renderer.h"
-#endif
-#ifdef __APPLE__
-#include "platform/metal_video_renderer.h"
-#endif
+#include "platform/video_renderer.h"
 #include "rd_log.h"
 #include "runtime/gui_runtime.h"
 
@@ -229,21 +224,14 @@ void GuiRuntime::CloseRemoteSession(std::shared_ptr<RemoteSession> props) {
     frame_snapshot = props->front_frame_;
     video_width = props->video_width_;
     video_height = props->video_height_;
-#if defined(__APPLE__) || defined(_WIN32)
     if ((!frame_snapshot || frame_snapshot->empty()) &&
         props->thumbnail_frame_ && !props->thumbnail_frame_->empty()) {
       frame_snapshot = props->thumbnail_frame_;
       video_width = props->thumbnail_width_;
       video_height = props->thumbnail_height_;
     }
-#endif
   }
-#if defined(__APPLE__) || defined(_WIN32)
-#if defined(_WIN32)
-  auto* native_renderer = windows_opengl_video_renderer_.get();
-#else
-  auto* native_renderer = mac_metal_video_renderer_.get();
-#endif
+  auto* native_renderer = video_renderer_.get();
   if ((!frame_snapshot || frame_snapshot->empty()) && native_renderer) {
     auto native_snapshot = std::make_shared<std::vector<unsigned char>>();
     if (native_renderer->CopyLatestNv12(props->remote_id_,
@@ -252,7 +240,6 @@ void GuiRuntime::CloseRemoteSession(std::shared_ptr<RemoteSession> props) {
       frame_snapshot = std::move(native_snapshot);
     }
   }
-#endif
 
   if (frame_snapshot && !frame_snapshot->empty() && video_width > 0 &&
       video_height > 0) {
@@ -276,12 +263,10 @@ void GuiRuntime::CloseRemoteSession(std::shared_ptr<RemoteSession> props) {
     }
   }
 
-#if defined(__APPLE__) || defined(_WIN32)
   if (native_renderer) {
     native_renderer->DiscardStream(props->remote_id_);
     video_frame_dirty_.store(true, std::memory_order_release);
   }
-#endif
 
   if (props->peer_) {
     LOG_INFO("[{}] Leave connection [{}]", props->local_id_, props->remote_id_);
@@ -343,12 +328,10 @@ void GuiRuntime::ResetRemoteSessionResources(
     std::lock_guard<std::mutex> lock(props->video_frame_mutex_);
     props->front_frame_.reset();
     props->back_frame_.reset();
-#if defined(__APPLE__) || defined(_WIN32)
     props->thumbnail_frame_.reset();
     props->thumbnail_width_ = 0;
     props->thumbnail_height_ = 0;
     props->background_snapshot_time_ = {};
-#endif
     props->video_width_ = 0;
     props->video_height_ = 0;
     props->video_size_ = 0;

src/gui/runtime/gui_runtime.cpp

@@ -12,12 +12,7 @@
 
 #include "display_stream_id.h"
 #include "localization.h"
-#if defined(_WIN32)
-#include "platform/windows_opengl_video_renderer.h"
-#endif
-#ifdef __APPLE__
-#include "platform/metal_video_renderer.h"
-#endif
+#include "platform/video_renderer.h"
 #include "rd_log.h"
 
 namespace crossdesk {

src/gui/runtime/gui_runtime.h

@@ -15,12 +15,7 @@
 
 namespace crossdesk {
 
-#if defined(_WIN32)
-class WindowsOpenGlVideoRenderer;
-#endif
-#ifdef __APPLE__
-class MacMetalVideoRenderer;
-#endif
+class VideoRenderer;
 
 // Shared GUI runtime. It owns subsystem controllers and cross-cutting session
 // state, but no window lifecycle, ImGui view, or transport callback methods.
@@ -90,12 +85,7 @@ class GuiRuntime : protected gui_detail::GuiState {
   KeyboardController keyboard_;
   PeerEventHandler peer_events_;
   std::atomic<bool> video_frame_dirty_{false};
-#if defined(_WIN32)
-  std::unique_ptr<WindowsOpenGlVideoRenderer> windows_opengl_video_renderer_;
-#endif
-#ifdef __APPLE__
-  std::unique_ptr<MacMetalVideoRenderer> mac_metal_video_renderer_;
-#endif
+  std::unique_ptr<VideoRenderer> video_renderer_;
 
  private:
   friend class ClipboardController;

src/gui/runtime/peer_event_handler.cpp

@@ -16,12 +16,7 @@
 #include "file_transfer.h"
 #include "localization.h"
 #include "platform.h"
-#if defined(_WIN32)
-#include "platform/windows_opengl_video_renderer.h"
-#endif
-#ifdef __APPLE__
-#include "platform/metal_video_renderer.h"
-#endif
+#include "platform/video_renderer.h"
 #include "rd_log.h"
 #include "runtime/gui_runtime.h"
 #include "runtime/remote_action_codec.h"
@@ -319,62 +314,46 @@ void PeerEventHandler::OnConnectionStatus(ConnectionStatus status,
         props->enable_mouse_control_ = false;
         runtime->ResetRemoteServiceStatus(*props);
 
-#if defined(__APPLE__) || defined(_WIN32)
         std::shared_ptr<std::vector<unsigned char>> native_snapshot;
         int native_snapshot_width = 0;
         int native_snapshot_height = 0;
         bool needs_native_snapshot = false;
         {
           std::lock_guard<std::mutex> lock(props->video_frame_mutex_);
-          needs_native_snapshot = !props->thumbnail_frame_ ||
-                                  props->thumbnail_frame_->empty();
+          needs_native_snapshot =
+              !props->thumbnail_frame_ || props->thumbnail_frame_->empty();
         }
         if (needs_native_snapshot) {
-#if defined(_WIN32)
-          auto* native_renderer =
-              runtime->windows_opengl_video_renderer_.get();
-#else
-          auto* native_renderer = runtime->mac_metal_video_renderer_.get();
-#endif
+          auto* native_renderer = runtime->video_renderer_.get();
           auto snapshot = std::make_shared<std::vector<unsigned char>>();
-          if (native_renderer && native_renderer->CopyLatestNv12(
-                                     remote_id, snapshot.get(),
-                                     &native_snapshot_width,
-                                     &native_snapshot_height)) {
+          if (native_renderer &&
+              native_renderer->CopyLatestNv12(remote_id, snapshot.get(),
+                                              &native_snapshot_width,
+                                              &native_snapshot_height)) {
             native_snapshot = std::move(snapshot);
           }
         }
-#endif
         {
           std::lock_guard<std::mutex> lock(props->video_frame_mutex_);
           props->front_frame_.reset();
           props->back_frame_.reset();
-#if defined(__APPLE__) || defined(_WIN32)
           if (native_snapshot &&
               (!props->thumbnail_frame_ || props->thumbnail_frame_->empty())) {
             props->thumbnail_frame_ = std::move(native_snapshot);
             props->thumbnail_width_ = native_snapshot_width;
             props->thumbnail_height_ = native_snapshot_height;
           }
-#endif
           props->video_width_ = 0;
           props->video_height_ = 0;
           props->video_size_ = 0;
           props->render_rect_dirty_ = true;
           props->stream_cleanup_pending_ = true;
         }
-#if defined(__APPLE__) || defined(_WIN32)
-#if defined(_WIN32)
-        auto* native_renderer = runtime->windows_opengl_video_renderer_.get();
-#else
-        auto* native_renderer = runtime->mac_metal_video_renderer_.get();
-#endif
+        auto* native_renderer = runtime->video_renderer_.get();
         if (native_renderer) {
           native_renderer->DiscardStream(remote_id);
-          runtime->video_frame_dirty_.store(true,
-                                            std::memory_order_release);
+          runtime->video_frame_dirty_.store(true, std::memory_order_release);
         }
-#endif
 
         runtime->focus_on_stream_window_ = false;
 

src/gui/runtime/peer_media_callbacks.cpp

@@ -1,36 +1,26 @@
-#include "runtime/peer_event_handler.h"
-
+#include <chrono>
 #include <cstring>
 #include <memory>
 #include <mutex>
 #include <string>
 #include <vector>
 
+#include "platform/video_renderer.h"
 #include "runtime/gui_runtime.h"
-#if defined(__APPLE__) || defined(_WIN32)
-#include <chrono>
-
-#if defined(_WIN32)
-#include "platform/windows_opengl_video_renderer.h"
-#else
-#include "platform/metal_video_renderer.h"
-#endif
-#endif
+#include "runtime/peer_event_handler.h"
 
 namespace crossdesk {
-#if defined(__APPLE__) || defined(_WIN32)
 namespace {
 
 constexpr auto kBackgroundSnapshotInterval = std::chrono::seconds(1);
 
 }  // namespace
-#endif
 
 void PeerEventHandler::OnReceiveVideoBuffer(
-    const XVideoFrame *video_frame, const char *user_id, size_t user_id_size,
-    const char *src_id, size_t src_id_size, void *user_data) {
-  auto *handler = static_cast<PeerEventHandler *>(user_data);
-  GuiRuntime *runtime = handler ? &handler->owner_ : nullptr;
+    const XVideoFrame* video_frame, const char* user_id, size_t user_id_size,
+    const char* src_id, size_t src_id_size, void* user_data) {
+  auto* handler = static_cast<PeerEventHandler*>(user_data);
+  GuiRuntime* runtime = handler ? &handler->owner_ : nullptr;
   if (!runtime) {
     return;
   }
@@ -41,32 +31,20 @@ void PeerEventHandler::OnReceiveVideoBuffer(
       runtime->remote_sessions_.end()) {
     return;
   }
-  GuiRuntime::RemoteSession *props =
+  GuiRuntime::RemoteSession* props =
       runtime->remote_sessions_.find(remote_id)->second.get();
 
   if (props->connection_established_) {
-#if defined(__APPLE__) || defined(_WIN32)
     bool background_snapshot_only = false;
-#if defined(_WIN32)
-    auto* native_renderer = runtime->windows_opengl_video_renderer_.get();
-    using NativeVideoRenderer = WindowsOpenGlVideoRenderer;
-#else
-    auto* native_renderer = runtime->mac_metal_video_renderer_.get();
-    using NativeVideoRenderer = MacMetalVideoRenderer;
-#endif
-    if (native_renderer && native_renderer->IsReady()
-#if defined(__APPLE__)
-        && native_renderer->IsAttached()
-#endif
-    ) {
+    auto* native_renderer = runtime->video_renderer_.get();
+    if (native_renderer && native_renderer->IsActive()) {
       const auto submit_result = native_renderer->SubmitNv12(
           remote_id, reinterpret_cast<const uint8_t*>(video_frame->data),
           video_frame->size, video_frame->width, video_frame->height);
-      if (submit_result == NativeVideoRenderer::SubmitResult::submitted) {
+      if (submit_result == VideoRenderer::SubmitResult::submitted) {
         std::lock_guard<std::mutex> lock(props->video_frame_mutex_);
-        const bool size_changed =
-            (props->video_width_ != video_frame->width) ||
-            (props->video_height_ != video_frame->height);
+        const bool size_changed = (props->video_width_ != video_frame->width) ||
+                                  (props->video_height_ != video_frame->height);
         if (size_changed) {
           props->render_rect_dirty_ = true;
         }
@@ -86,22 +64,19 @@ void PeerEventHandler::OnReceiveVideoBuffer(
         runtime->video_frame_dirty_.store(true, std::memory_order_release);
         return;
       }
-      if (submit_result == NativeVideoRenderer::SubmitResult::dropped ||
-          submit_result == NativeVideoRenderer::SubmitResult::failed) {
+      if (submit_result == VideoRenderer::SubmitResult::dropped ||
+          submit_result == VideoRenderer::SubmitResult::failed) {
         // Keep presenting the last native frame. A later decoded frame can
         // reuse the renderer without changing ownership mid-window.
         props->streaming_ = true;
         return;
       }
-      if (submit_result ==
-          NativeVideoRenderer::SubmitResult::not_selected) {
+      if (submit_result == VideoRenderer::SubmitResult::not_selected) {
         background_snapshot_only = true;
       }
     }
-#endif
     {
       std::lock_guard<std::mutex> lock(props->video_frame_mutex_);
-#if defined(__APPLE__) || defined(_WIN32)
       const auto now = std::chrono::steady_clock::now();
       if (background_snapshot_only && props->thumbnail_frame_ &&
           !props->thumbnail_frame_->empty() &&
@@ -112,7 +87,6 @@ void PeerEventHandler::OnReceiveVideoBuffer(
         props->streaming_ = true;
         return;
       }
-#endif
 
       // Allocate a third buffer only while the UI still owns the old snapshot.
       if (!props->back_frame_ || props->back_frame_.use_count() != 1) {
@@ -137,32 +111,28 @@ void PeerEventHandler::OnReceiveVideoBuffer(
       props->video_size_ = video_frame->size;
 
       props->front_frame_.swap(props->back_frame_);
-#if defined(__APPLE__) || defined(_WIN32)
       props->thumbnail_frame_ = props->front_frame_;
       props->thumbnail_width_ = video_frame->width;
       props->thumbnail_height_ = video_frame->height;
       if (background_snapshot_only) {
         props->background_snapshot_time_ = now;
       }
-#endif
       ++props->video_frame_sequence_;
     }
 
     props->streaming_ = true;
-#if defined(__APPLE__) || defined(_WIN32)
     if (background_snapshot_only) {
       return;
     }
-#endif
     runtime->video_frame_dirty_.store(true, std::memory_order_release);
   }
 }
 
 void PeerEventHandler::OnReceiveAudioBuffer(
-    const char *data, size_t size, const char *user_id, size_t user_id_size,
-    const char *src_id, size_t src_id_size, void *user_data) {
-  auto *handler = static_cast<PeerEventHandler *>(user_data);
-  GuiRuntime *runtime = handler ? &handler->owner_ : nullptr;
+    const char* data, size_t size, const char* user_id, size_t user_id_size,
+    const char* src_id, size_t src_id_size, void* user_data) {
+  auto* handler = static_cast<PeerEventHandler*>(user_data);
+  GuiRuntime* runtime = handler ? &handler->owner_ : nullptr;
   if (!runtime) {
     return;
   }
@@ -172,4 +142,4 @@ void PeerEventHandler::OnReceiveAudioBuffer(
   runtime->devices_.PushAudio(data, size);
 }
 
-} // namespace crossdesk
+}  // namespace crossdesk

src/gui/runtime/remote_session.h

@@ -110,14 +110,12 @@ struct RemoteSession {
   std::mutex video_frame_mutex_;
   std::shared_ptr<std::vector<unsigned char>> front_frame_;
   std::shared_ptr<std::vector<unsigned char>> back_frame_;
-#if defined(__APPLE__) || defined(_WIN32)
   // Retains the most recent background CPU frame across disconnect cleanup
   // so closing an unselected tab can still update its recent-item thumbnail.
   std::shared_ptr<std::vector<unsigned char>> thumbnail_frame_;
   int thumbnail_width_ = 0;
   int thumbnail_height_ = 0;
   std::chrono::steady_clock::time_point background_snapshot_time_;
-#endif
   bool render_rect_dirty_ = false;
   bool stream_cleanup_pending_ = false;
   float mouse_pos_x_ = 0;

xmake/targets.lua

@@ -270,6 +270,7 @@ function setup_targets()
             "device_controller", "thumbnail", "version_checker", "tools")
         add_files("src/gui/render.cpp", "src/gui/application/gui_application.cpp",
             "src/gui/application/portable_service_integration.cpp",
+            "src/gui/rendering/*.cpp",
             "src/gui/runtime/*.cpp",
             "src/gui/features/devices/*.cpp", "src/gui/features/input/*.cpp",
             "src/gui/features/clipboard/*.cpp", "src/gui/features/file_transfer/*.cpp",
@@ -279,15 +280,19 @@ function setup_targets()
             add_cxxflags("/bigobj")
             add_links("opengl32")
             add_files("src/gui/platform/tray/win_tray.cpp",
-                "src/gui/platform/windows_opengl_video_renderer.cpp")
+                "src/gui/platform/opengl_video_renderer.cpp",
+                "src/gui/platform/video_renderer_factory_opengl.cpp")
             add_includedirs("src/service/windows", {public = true})
         elseif is_os("macosx") then
             add_files("src/gui/runtime/*.mm", "src/gui/platform/tray/*.mm",
                 "src/gui/platform/window_drag_mac.mm",
-                "src/gui/platform/metal_video_renderer.mm")
+                "src/gui/platform/metal_video_renderer.mm",
+                "src/gui/platform/video_renderer_factory_metal.mm")
         elseif is_os("linux") then
             add_links("GL")
-            add_files("src/gui/platform/tray/linux_tray.cpp")
+            add_files("src/gui/platform/tray/linux_tray.cpp",
+                "src/gui/platform/opengl_video_renderer.cpp",
+                "src/gui/platform/video_renderer_factory_opengl.cpp")
         end
 
     if is_os("windows") then