GPU: Fix various capture/screenshot issues

- Simplify functions for rendering display.

- Post-processing should only when the target size matches the window.
  Otherwise the shaders are constantly recompiled.

- Include border overlay in capture/screenshots when above condition
  is satisfied.

- Relax video alignment size when using screen resolution + auto.

- Fix "Internal Resolution (Uncorrected)" capture mode.

src/core/gpu_backend.cpp

@@ -653,25 +653,39 @@ void GPUBackend::UpdateStatistics(u32 frame_count)
   ResetStatistics();
 }
 
-bool GPUBackend::RenderScreenshotToBuffer(u32 width, u32 height, bool postfx, Image* out_image)
+bool GPUBackend::RenderScreenshotToBuffer(u32 width, u32 height, bool postfx, bool apply_aspect_ratio, Image* out_image,
+                                          Error* error)
 {
   bool result;
   GPUThread::RunOnBackend(
-    [width, height, postfx, out_image, &result](GPUBackend* backend) {
+    [width, height, postfx, apply_aspect_ratio, out_image, error, &result](GPUBackend* backend) {
       if (!backend)
+      {
+        Error::SetStringView(error, "No GPU backend.");
+        result = false;
         return;
+      }
 
-      GSVector4i draw_rect, display_rect;
+      // Post-processing requires that the size match the window.
-      backend->m_presenter.CalculateDrawRect(static_cast<s32>(width), static_cast<s32>(height), true, true,
+      const bool really_postfx = postfx && g_gpu_device->HasMainSwapChain();
-                                             &display_rect, &draw_rect);
+      u32 image_width, image_height;
+      if (really_postfx)
+      {
+        image_width = g_gpu_device->GetMainSwapChain()->GetWidth();
+        image_height = g_gpu_device->GetMainSwapChain()->GetHeight();
+      }
+      else
+      {
+        // Crop it if border overlay isn't enabled.
+        GSVector4i draw_rect, display_rect;
+        backend->GetPresenter().CalculateDrawRect(static_cast<s32>(width), static_cast<s32>(height), apply_aspect_ratio,
+                                                  &display_rect, &draw_rect);
+        image_width = static_cast<u32>(display_rect.width());
+        image_height = static_cast<u32>(display_rect.height());
+      }
 
-      // Crop it.
+      result = backend->GetPresenter().RenderScreenshotToBuffer(image_width, image_height, really_postfx,
-      const u32 cropped_width = static_cast<u32>(display_rect.width());
+                                                                apply_aspect_ratio, out_image, error);
-      const u32 cropped_height = static_cast<u32>(display_rect.height());
-      draw_rect = draw_rect.sub32(display_rect.xyxy());
-      display_rect = display_rect.sub32(display_rect.xyxy());
-      result = backend->m_presenter.RenderScreenshotToBuffer(cropped_width, cropped_height, display_rect, draw_rect,
-                                                             postfx, out_image);
       backend->RestoreDeviceContext();
     },
     true, false);
@@ -687,19 +701,29 @@ void GPUBackend::RenderScreenshotToFile(const std::string_view path, DisplayScre
       if (!backend)
         return;
 
-      u32 width, height;
+      const GSVector2i size = backend->GetPresenter().CalculateScreenshotSize(mode);
-      GSVector4i display_rect, draw_rect;
+      if (size.x == 0 || size.y == 0)
-      backend->m_presenter.CalculateScreenshotSize(mode, &width, &height, &display_rect, &draw_rect);
-
-      const bool internal_resolution = (mode != DisplayScreenshotMode::ScreenResolution);
-      if (width == 0 || height == 0)
         return;
 
+      std::string osd_key;
+      if (show_osd_message)
+        osd_key = fmt::format("ScreenshotSaver_{}", path);
+
+      const bool internal_resolution = (mode != DisplayScreenshotMode::ScreenResolution);
+      const bool apply_aspect_ratio = (mode != DisplayScreenshotMode::UncorrectedInternalResolution);
+      Error error;
       Image image;
-      if (!backend->m_presenter.RenderScreenshotToBuffer(width, height, display_rect, draw_rect, !internal_resolution,
+      if (!backend->m_presenter.RenderScreenshotToBuffer(size.x, size.y, !internal_resolution, apply_aspect_ratio,
-                                                         &image))
+                                                         &image, &error))
       {
-        ERROR_LOG("Failed to render {}x{} screenshot", width, height);
+        ERROR_LOG("Failed to render {}x{} screenshot: {}", size.x, size.y, error.GetDescription());
+        if (show_osd_message)
+        {
+          Host::AddIconOSDWarning(
+            std::move(osd_key), ICON_EMOJI_WARNING,
+            fmt::format(TRANSLATE_FS("GPU", "Failed to save screenshot:\n{}"), error.GetDescription()));
+        }
+
         backend->RestoreDeviceContext();
         return;
       }
@@ -707,19 +731,23 @@ void GPUBackend::RenderScreenshotToFile(const std::string_view path, DisplayScre
       // no more GPU calls
       backend->RestoreDeviceContext();
 
-      Error error;
       auto fp = FileSystem::OpenManagedCFile(path.c_str(), "wb", &error);
       if (!fp)
       {
         ERROR_LOG("Can't open file '{}': {}", Path::GetFileName(path), error.GetDescription());
+        if (show_osd_message)
+        {
+          Host::AddIconOSDWarning(
+            std::move(osd_key), ICON_EMOJI_WARNING,
+            fmt::format(TRANSLATE_FS("GPU", "Failed to save screenshot:\n{}"), error.GetDescription()));
+        }
+
         return;
       }
 
-      std::string osd_key;
       if (show_osd_message)
       {
         // Use a 60 second timeout to give it plenty of time to actually save.
-        osd_key = fmt::format("ScreenshotSaver_{}", path);
         Host::AddIconOSDMessage(osd_key, ICON_EMOJI_CAMERA_WITH_FLASH,
                                 fmt::format(TRANSLATE_FS("GPU", "Saving screenshot to '{}'."), Path::GetFileName(path)),
                                 60.0f);

src/core/gpu_backend.h

@@ -58,7 +58,8 @@ public:
   static std::unique_ptr<GPUBackend> CreateSoftwareBackend(GPUPresenter& presenter);
   static std::unique_ptr<GPUBackend> CreateNullBackend(GPUPresenter& presenter);
 
-  static bool RenderScreenshotToBuffer(u32 width, u32 height, bool postfx, Image* out_image);
+  static bool RenderScreenshotToBuffer(u32 width, u32 height, bool postfx, bool apply_aspect_ratio, Image* out_image,
+                                       Error* error);
   static void RenderScreenshotToFile(const std::string_view path, DisplayScreenshotMode mode, u8 quality,
                                      bool show_osd_message);
 

src/core/gpu_hw.cpp

@@ -746,8 +746,7 @@ u32 GPU_HW::CalculateResolutionScale() const
     {
       GSVector4i display_rect, draw_rect;
       m_presenter.CalculateDrawRect(g_gpu_device->GetMainSwapChain()->GetWidth(),
-                                    g_gpu_device->GetMainSwapChain()->GetHeight(), true, true, &display_rect,
+                                    g_gpu_device->GetMainSwapChain()->GetHeight(), true, &display_rect, &draw_rect);
-                                    &draw_rect);
 
       // We use the draw rect to determine scaling. This way we match the resolution as best we can, regardless of the
       // anamorphic aspect ratio.

src/core/gpu_presenter.cpp

@@ -377,65 +377,60 @@ void GPUPresenter::SetDisplayTexture(GPUTexture* texture, s32 view_x, s32 view_y
   m_display_texture_view_height = view_height;
 }
 
-GPUDevice::PresentResult GPUPresenter::PresentDisplay()
+GPUDevice::PresentResult GPUPresenter::RenderDisplay(GPUTexture* target, const GSVector2i target_size, bool postfx,
+                                                     bool apply_aspect_ratio)
 {
-  DebugAssert(g_gpu_device->HasMainSwapChain());
+  GL_SCOPE_FMT("RenderDisplay: {}x{}", target_size.x, target_size.y);
 
-  u32 display_area_width = g_gpu_device->GetMainSwapChain()->GetWidth();
+  if (m_display_texture)
-  u32 display_area_height = g_gpu_device->GetMainSwapChain()->GetHeight();
+    m_display_texture->MakeReadyForSampling();
+
+  DebugAssert(target || g_gpu_device->HasMainSwapChain());
+  DebugAssert(!postfx || target_size.eq(g_gpu_device->GetMainSwapChain()->GetSizeVec()));
+
+  GPUSwapChain* const swap_chain = g_gpu_device->GetMainSwapChain();
+  const WindowInfo::PreRotation prerotation = target ? WindowInfo::PreRotation::Identity : swap_chain->GetPreRotation();
+  const bool is_vram_view = g_gpu_settings.gpu_show_vram;
+  const bool have_overlay = (postfx && !is_vram_view && HasBorderOverlay());
+  const bool have_prerotation = (prerotation != WindowInfo::PreRotation::Identity);
+  GL_INS(have_overlay ? "Overlay is ENABLED" : "Overlay is disabled");
+  GL_INS_FMT("Prerotation: {}", static_cast<u32>(prerotation));
+  GL_INS_FMT("Final target size: {}x{}", target_size.x, target_size.y);
+
+  // Compute draw area.
+  GSVector4i display_rect;
+  GSVector4i draw_rect;
   GSVector4i overlay_display_rect = GSVector4i::zero();
   GSVector4i overlay_rect = GSVector4i::zero();
-  if (m_border_overlay_texture)
+  if (have_overlay)
   {
-    overlay_rect = GSVector4i::rfit(GSVector4i(0, 0, display_area_width, display_area_height),
+    overlay_rect = GSVector4i::rfit(GSVector4i::loadh(target_size), m_border_overlay_texture->GetSizeVec());
-                                    m_border_overlay_texture->GetSizeVec());
 
     const GSVector2 scale = GSVector2(overlay_rect.rsize()) / GSVector2(m_border_overlay_texture->GetSizeVec());
     overlay_display_rect =
       GSVector4i(GSVector4(m_border_overlay_display_rect) * GSVector4::xyxy(scale)).add32(overlay_rect.xyxy());
-    display_area_width = overlay_display_rect.width();
+
-    display_area_height = overlay_display_rect.height();
+    // Draw to the overlay area instead of the whole screen.
+    CalculateDrawRect(overlay_display_rect.width(), overlay_display_rect.height(), apply_aspect_ratio, &display_rect,
+                      &draw_rect);
+
+    // Apply overlay area offset.
+    display_rect = display_rect.add32(overlay_display_rect.xyxy());
+    draw_rect = draw_rect.add32(overlay_display_rect.xyxy());
+  }
+  else
+  {
+    CalculateDrawRect(target_size.x, target_size.y, apply_aspect_ratio, &display_rect, &draw_rect);
   }
-
-  GSVector4i display_rect;
-  GSVector4i draw_rect;
-  CalculateDrawRect(display_area_width, display_area_height, !g_gpu_settings.gpu_show_vram, true, &display_rect,
-                    &draw_rect);
-
-  display_rect = display_rect.add32(overlay_display_rect.xyxy());
-  draw_rect = draw_rect.add32(overlay_display_rect.xyxy());
-
-  return RenderDisplay(nullptr, overlay_rect, overlay_display_rect, display_rect, draw_rect,
-                       !g_gpu_settings.gpu_show_vram);
-}
-
-GPUDevice::PresentResult GPUPresenter::RenderDisplay(GPUTexture* target, const GSVector4i overlay_rect,
-                                                     const GSVector4i overlay_display_rect,
-                                                     const GSVector4i display_rect, const GSVector4i draw_rect,
-                                                     bool postfx)
-{
-  GL_SCOPE_FMT("RenderDisplay: {}", draw_rect);
-
-  if (m_display_texture)
-    m_display_texture->MakeReadyForSampling();
 
   // There's a bunch of scenarios where we need to use intermediate buffers.
   // If we have post-processing and overlays enabled, postfx needs to happen on an intermediate buffer first.
   // If pre-rotation is enabled with post-processing, we need to draw to an intermediate buffer, and apply the
   // rotation at the end.
-  GPUSwapChain* const swap_chain = g_gpu_device->GetMainSwapChain();
-  const WindowInfo::PreRotation prerotation = target ? WindowInfo::PreRotation::Identity : swap_chain->GetPreRotation();
-  const bool have_overlay = static_cast<bool>(m_border_overlay_texture);
-  const bool have_prerotation = (prerotation != WindowInfo::PreRotation::Identity);
-  const GSVector2i target_size = target ? target->GetSizeVec() : swap_chain->GetSizeVec();
-  GL_INS(have_overlay ? "Overlay is ENABLED" : "Overlay is disabled");
-  GL_INS_FMT("Prerotation: {}", static_cast<u32>(prerotation));
-  GL_INS_FMT("Final target size: {}x{}", target_size.x, target_size.y);
-
-  // Postfx active?
   const GSVector2i postfx_size = have_overlay ? overlay_display_rect.rsize() : target_size;
-  const bool really_postfx = (postfx && m_display_postfx && m_display_postfx->IsActive() && m_display_postfx &&
+  const bool really_postfx =
-                              m_display_postfx->CheckTargets(m_present_format, postfx_size.x, postfx_size.y));
+    (postfx && !is_vram_view && m_display_postfx && m_display_postfx->IsActive() && m_display_postfx &&
+     m_display_postfx->CheckTargets(m_present_format, postfx_size.x, postfx_size.y));
   GL_INS(really_postfx ? "Post-processing is ENABLED" : "Post-processing is disabled");
   GL_INS_FMT("Post-processing render target size: {}x{}", postfx_size.x, postfx_size.y);
 
@@ -708,16 +703,11 @@ void GPUPresenter::SendDisplayToMediaCapture(MediaCapture* cap)
 
   const bool apply_aspect_ratio =
     (g_gpu_settings.display_screenshot_mode != DisplayScreenshotMode::UncorrectedInternalResolution);
-  const bool postfx = (g_gpu_settings.display_screenshot_mode != DisplayScreenshotMode::InternalResolution);
+  const bool postfx =
-  GSVector4i display_rect, draw_rect;
+    (g_gpu_settings.display_screenshot_mode == DisplayScreenshotMode::ScreenResolution &&
-  CalculateDrawRect(target->GetWidth(), target->GetHeight(), !g_gpu_settings.gpu_show_vram, apply_aspect_ratio,
+     g_gpu_device->HasMainSwapChain() && target->GetSizeVec().eq(g_gpu_device->GetMainSwapChain()->GetSizeVec()));
-                    &display_rect, &draw_rect);
 
-  // Not cleared by RenderDisplay().
+  if (RenderDisplay(target, target->GetSizeVec(), postfx, apply_aspect_ratio) != GPUDevice::PresentResult::OK ||
-  g_gpu_device->ClearRenderTarget(target, GPUDevice::DEFAULT_CLEAR_COLOR);
-
-  if (RenderDisplay(target, GSVector4i::zero(), GSVector4i::zero(), display_rect, draw_rect, postfx) !=
-        GPUDevice::PresentResult::OK ||
       !cap->DeliverVideoFrame(target)) [[unlikely]]
   {
     WARNING_LOG("Failed to render/deliver video capture frame.");
@@ -907,7 +897,7 @@ bool GPUPresenter::ApplyChromaSmoothing()
   return true;
 }
 
-void GPUPresenter::CalculateDrawRect(s32 window_width, s32 window_height, bool apply_rotation, bool apply_aspect_ratio,
+void GPUPresenter::CalculateDrawRect(s32 window_width, s32 window_height, bool apply_aspect_ratio,
                                      GSVector4i* display_rect, GSVector4i* draw_rect) const
 {
   const bool integer_scale = (g_gpu_settings.display_scaling == DisplayScalingMode::NearestInteger ||
@@ -919,9 +909,10 @@ void GPUPresenter::CalculateDrawRect(s32 window_width, s32 window_height, bool a
   const s32 display_origin_top = show_vram ? 0 : m_display_origin_top;
   const u32 display_vram_width = show_vram ? VRAM_WIDTH : m_display_vram_width;
   const u32 display_vram_height = show_vram ? VRAM_HEIGHT : m_display_vram_height;
-  const float display_pixel_aspect_ratio = show_vram ? 1.0f : m_display_pixel_aspect_ratio;
+  const float display_pixel_aspect_ratio = (show_vram || !apply_aspect_ratio) ? 1.0f : m_display_pixel_aspect_ratio;
+  const DisplayRotation display_rotation = show_vram ? DisplayRotation::Normal : g_gpu_settings.display_rotation;
   GPU::CalculateDrawRect(window_width, window_height, display_width, display_height, display_origin_left,
-                         display_origin_top, display_vram_width, display_vram_height, g_gpu_settings.display_rotation,
+                         display_origin_top, display_vram_width, display_vram_height, display_rotation,
                          g_gpu_settings.display_alignment, display_pixel_aspect_ratio,
                          g_gpu_settings.display_stretch_vertically, integer_scale, display_rect, draw_rect);
 }
@@ -954,16 +945,17 @@ bool GPUPresenter::PresentFrame(GPUPresenter* presenter, GPUBackend* backend, bo
     ImGuiManager::RenderDebugWindows();
   }
 
+  GPUSwapChain* const swap_chain = g_gpu_device->GetMainSwapChain();
   const GPUDevice::PresentResult pres =
-    skip_present ?
+    skip_present ? GPUDevice::PresentResult::SkipPresent :
-      GPUDevice::PresentResult::SkipPresent :
+                   (presenter ? presenter->RenderDisplay(nullptr, swap_chain->GetSizeVec(), true, true) :
-      (presenter ? presenter->PresentDisplay() : g_gpu_device->BeginPresent(g_gpu_device->GetMainSwapChain()));
+                                g_gpu_device->BeginPresent(swap_chain));
   if (pres == GPUDevice::PresentResult::OK)
   {
     if (presenter)
       presenter->m_skipped_present_count = 0;
 
-    g_gpu_device->RenderImGui(g_gpu_device->GetMainSwapChain());
+    g_gpu_device->RenderImGui(swap_chain);
 
     const GPUDevice::Features features = g_gpu_device->GetFeatures();
     const bool scheduled_present = (present_time != 0);
@@ -977,7 +969,7 @@ bool GPUPresenter::PresentFrame(GPUPresenter* presenter, GPUBackend* backend, bo
       SleepUntilPresentTime(present_time);
     }
 
-    g_gpu_device->EndPresent(g_gpu_device->GetMainSwapChain(), explicit_present, timed_present ? present_time : 0);
+    g_gpu_device->EndPresent(swap_chain, explicit_present, timed_present ? present_time : 0);
 
     if (g_gpu_device->IsGPUTimingEnabled())
       PerformanceCounters::AccumulateGPUTime();
@@ -985,7 +977,7 @@ bool GPUPresenter::PresentFrame(GPUPresenter* presenter, GPUBackend* backend, bo
     if (explicit_present)
     {
       SleepUntilPresentTime(present_time);
-      g_gpu_device->SubmitPresent(g_gpu_device->GetMainSwapChain());
+      g_gpu_device->SubmitPresent(swap_chain);
     }
   }
   else
@@ -1029,8 +1021,8 @@ void GPUPresenter::SleepUntilPresentTime(u64 present_time)
 #endif
 }
 
-bool GPUPresenter::RenderScreenshotToBuffer(u32 width, u32 height, const GSVector4i display_rect,
+bool GPUPresenter::RenderScreenshotToBuffer(u32 width, u32 height, bool postfx, bool apply_aspect_ratio,
-                                            const GSVector4i draw_rect, bool postfx, Image* out_image)
+                                            Image* out_image, Error* error)
 {
   const ImageFormat image_format = GPUTexture::GetImageFormatForTextureFormat(m_present_format);
   if (image_format == ImageFormat::None)
@@ -1042,24 +1034,26 @@ bool GPUPresenter::RenderScreenshotToBuffer(u32 width, u32 height, const GSVecto
     return false;
 
   g_gpu_device->ClearRenderTarget(render_texture.get(), GPUDevice::DEFAULT_CLEAR_COLOR);
-
+  if (RenderDisplay(render_texture.get(), render_texture->GetSizeVec(), postfx, apply_aspect_ratio) !=
-  // TODO: this should use copy shader instead.
+      GPUDevice::PresentResult::OK)
-  RenderDisplay(render_texture.get(), GSVector4i::zero(), GSVector4i::zero(), display_rect, draw_rect, postfx);
+  {
+    Error::SetStringView(error, "RenderDisplay() failed");
+    return false;
+  }
 
   Image image(width, height, image_format);
 
-  Error error;
   std::unique_ptr<GPUDownloadTexture> dltex;
   if (g_gpu_device->GetFeatures().memory_import)
   {
     dltex = g_gpu_device->CreateDownloadTexture(width, height, m_present_format, image.GetPixels(),
-                                                image.GetStorageSize(), image.GetPitch(), &error);
+                                                image.GetStorageSize(), image.GetPitch(), error);
   }
   if (!dltex)
   {
-    if (!(dltex = g_gpu_device->CreateDownloadTexture(width, height, m_present_format, &error)))
+    if (!(dltex = g_gpu_device->CreateDownloadTexture(width, height, m_present_format, error)))
     {
-      ERROR_LOG("Failed to create {}x{} download texture: {}", width, height, error.GetDescription());
+      Error::AddPrefixFmt(error, "Failed to create {}x{} download texture: ", width, height);
       return false;
     }
   }
@@ -1072,8 +1066,7 @@ bool GPUPresenter::RenderScreenshotToBuffer(u32 width, u32 height, const GSVecto
   return true;
 }
 
-void GPUPresenter::CalculateScreenshotSize(DisplayScreenshotMode mode, u32* width, u32* height,
+GSVector2i GPUPresenter::CalculateScreenshotSize(DisplayScreenshotMode mode) const
-                                           GSVector4i* display_rect, GSVector4i* draw_rect) const
 {
   const bool internal_resolution = (mode != DisplayScreenshotMode::ScreenResolution || g_gpu_settings.gpu_show_vram);
   if (internal_resolution && m_display_texture_view_width != 0 && m_display_texture_view_height != 0)
@@ -1098,24 +1091,16 @@ void GPUPresenter::CalculateScreenshotSize(DisplayScreenshotMode mode, u32* widt
         f_width = f_width / (f_height / max_texture_size);
       }
 
-      *width = static_cast<u32>(std::ceil(f_width));
+      return GSVector2i(static_cast<s32>(std::ceil(f_width)), static_cast<s32>(std::ceil(f_height)));
-      *height = static_cast<u32>(std::ceil(f_height));
     }
     else // if (mode == DisplayScreenshotMode::UncorrectedInternalResolution)
     {
-      *width = m_display_texture_view_width;
+      return GSVector2i(m_display_texture_view_width, m_display_texture_view_height);
-      *height = m_display_texture_view_height;
     }
-
-    // Remove padding, it's not part of the framebuffer.
-    *draw_rect = GSVector4i(0, 0, static_cast<s32>(*width), static_cast<s32>(*height));
-    *display_rect = *draw_rect;
   }
   else
   {
-    *width = g_gpu_device->HasMainSwapChain() ? g_gpu_device->GetMainSwapChain()->GetWidth() : 1;
+    return g_gpu_device->HasMainSwapChain() ? g_gpu_device->GetMainSwapChain()->GetSizeVec() : GSVector2i(1, 1);
-    *height = g_gpu_device->HasMainSwapChain() ? g_gpu_device->GetMainSwapChain()->GetHeight() : 1;
-    CalculateDrawRect(*width, *height, true, !g_settings.gpu_show_vram, display_rect, draw_rect);
   }
 }
 

src/core/gpu_presenter.h

@@ -44,6 +44,7 @@ public:
   ALWAYS_INLINE s32 GetDisplayTextureViewHeight() const { return m_display_texture_view_height; }
   ALWAYS_INLINE GPUTexture* GetDisplayTexture() const { return m_display_texture; }
   ALWAYS_INLINE bool HasDisplayTexture() const { return m_display_texture; }
+  ALWAYS_INLINE bool HasBorderOverlay() const { return static_cast<bool>(m_border_overlay_texture); }
 
   bool Initialize(Error* error);
 
@@ -59,16 +60,15 @@ public:
   bool ApplyChromaSmoothing();
 
   /// Helper function for computing the draw rectangle in a larger window.
-  void CalculateDrawRect(s32 window_width, s32 window_height, bool apply_rotation, bool apply_aspect_ratio,
+  void CalculateDrawRect(s32 window_width, s32 window_height, bool apply_aspect_ratio, GSVector4i* display_rect,
-                         GSVector4i* display_rect, GSVector4i* draw_rect) const;
+                         GSVector4i* draw_rect) const;
 
   /// Helper function for computing screenshot bounds.
-  void CalculateScreenshotSize(DisplayScreenshotMode mode, u32* width, u32* height, GSVector4i* display_rect,
+  GSVector2i CalculateScreenshotSize(DisplayScreenshotMode mode) const;
-                               GSVector4i* draw_rect) const;
 
   /// Renders the display, optionally with postprocessing to the specified image.
-  bool RenderScreenshotToBuffer(u32 width, u32 height, const GSVector4i display_rect, const GSVector4i draw_rect,
+  bool RenderScreenshotToBuffer(u32 width, u32 height, bool postfx, bool apply_aspect_ratio, Image* out_image,
-                                bool postfx, Image* out_image);
+                                Error* error);
 
   /// Sends the current frame to media capture.
   void SendDisplayToMediaCapture(MediaCapture* cap);
@@ -98,14 +98,10 @@ private:
 
   static void SleepUntilPresentTime(u64 present_time);
 
-  /// Draws the current display texture, with any post-processing.
-  GPUDevice::PresentResult PresentDisplay();
-
   bool CompileDisplayPipelines(bool display, bool deinterlace, bool chroma_smoothing, Error* error);
 
-  GPUDevice::PresentResult RenderDisplay(GPUTexture* target, const GSVector4i overlay_rect,
+  GPUDevice::PresentResult RenderDisplay(GPUTexture* target, const GSVector2i target_size, bool postfx,
-                                         const GSVector4i overlay_display_rect, const GSVector4i display_rect,
+                                         bool apply_aspect_ratio);
-                                         const GSVector4i draw_rect, bool postfx);
 
   void DrawDisplay(const GSVector2i target_size, const GSVector4i display_rect, bool dst_alpha_blend,
                    DisplayRotation rotation, WindowInfo::PreRotation prerotation);

src/core/system.cpp

@@ -3257,7 +3257,9 @@ bool System::SaveStateToBuffer(SaveStateBuffer* buffer, Error* error, u32 screen
   // save screenshot
   if (screenshot_size > 0)
   {
-    if (GPUBackend::RenderScreenshotToBuffer(screenshot_size, screenshot_size, false, &buffer->screenshot))
+    Error screenshot_error;
+    if (GPUBackend::RenderScreenshotToBuffer(screenshot_size, screenshot_size, false, true, &buffer->screenshot,
+                                             &screenshot_error))
     {
       if (g_gpu_device->UsesLowerLeftOrigin())
         buffer->screenshot.FlipY();
@@ -3265,12 +3267,11 @@ bool System::SaveStateToBuffer(SaveStateBuffer* buffer, Error* error, u32 screen
       // Ensure it's RGBA8.
       if (buffer->screenshot.GetFormat() != ImageFormat::RGBA8)
       {
-        Error convert_error;
+        std::optional<Image> screenshot_rgba8 = buffer->screenshot.ConvertToRGBA8(&screenshot_error);
-        std::optional<Image> screenshot_rgba8 = buffer->screenshot.ConvertToRGBA8(&convert_error);
         if (!screenshot_rgba8.has_value())
         {
           ERROR_LOG("Failed to convert {} screenshot to RGBA8: {}",
-                    Image::GetFormatName(buffer->screenshot.GetFormat()), convert_error.GetDescription());
+                    Image::GetFormatName(buffer->screenshot.GetFormat()), screenshot_error.GetDescription());
           buffer->screenshot.Invalidate();
         }
         else
@@ -3281,8 +3282,8 @@ bool System::SaveStateToBuffer(SaveStateBuffer* buffer, Error* error, u32 screen
     }
     else
     {
-      WARNING_LOG("Failed to save {}x{} screenshot for save state due to render/conversion failure", screenshot_size,
+      WARNING_LOG("Failed to save {}x{} screenshot for save state: {}", screenshot_size, screenshot_size,
-                  screenshot_size);
+                  screenshot_error.GetDescription());
     }
   }
 
@@ -5347,14 +5348,16 @@ bool System::StartMediaCapture(std::string path)
   {
     // need to query this on the GPU thread
     GPUThread::RunOnBackend(
-      [path = std::move(path), capture_audio](GPUBackend* backend) mutable {
+      [path = std::move(path), capture_audio, mode = g_settings.display_screenshot_mode](GPUBackend* backend) mutable {
         if (!backend)
           return;
 
-        GSVector4i unused_display_rect, unused_draw_rect;
+        // Prefer aligning for non-window size.
-        u32 video_width, video_height;
+        const GSVector2i video_size = backend->GetPresenter().CalculateScreenshotSize(mode);
-        backend->GetPresenter().CalculateScreenshotSize(DisplayScreenshotMode::InternalResolution, &video_width,
+        u32 video_width = static_cast<u32>(video_size.x);
-                                                        &video_height, &unused_display_rect, &unused_draw_rect);
+        u32 video_height = static_cast<u32>(video_size.y);
+        if (mode != DisplayScreenshotMode::ScreenResolution)
+          MediaCapture::AdjustVideoSize(&video_width, &video_height);
 
         // fire back to the CPU thread to actually start the capture
         Host::RunOnCPUThread([path = std::move(path), capture_audio, video_width, video_height]() mutable {
@@ -5369,6 +5372,7 @@ bool System::StartMediaCapture(std::string path)
     Host::GetUIntSettingValue("MediaCapture", "VideoWidth", Settings::DEFAULT_MEDIA_CAPTURE_VIDEO_WIDTH);
   u32 video_height =
     Host::GetUIntSettingValue("MediaCapture", "VideoHeight", Settings::DEFAULT_MEDIA_CAPTURE_VIDEO_HEIGHT);
+  MediaCapture::AdjustVideoSize(&video_width, &video_height);
 
   return StartMediaCapture(std::move(path), capture_video, capture_audio, video_width, video_height);
 }
@@ -5385,8 +5389,6 @@ bool System::StartMediaCapture(std::string path, bool capture_video, bool captur
   const WindowInfo& main_window_info = GPUThread::GetRenderWindowInfo();
   const GPUTexture::Format capture_format =
     main_window_info.IsSurfaceless() ? GPUTexture::Format::RGBA8 : main_window_info.surface_format;
-  if (capture_video)
-    MediaCapture::AdjustVideoSize(&video_width, &video_height);
 
   // TODO: Render anamorphic capture instead?
   constexpr float aspect = 1.0f;

src/util/media_capture.cpp

@@ -200,9 +200,7 @@ bool MediaCaptureBase::BeginCapture(float fps, float aspect, u32 width, u32 heig
     Error::SetStringView(error, "No path specified.");
     return false;
   }
-  else if (capture_video &&
+  else if (capture_video && (fps == 0.0f || m_video_width == 0 || m_video_height == 0))
-           (fps == 0.0f || m_video_width == 0 || !Common::IsAlignedPow2(m_video_width, VIDEO_WIDTH_ALIGNMENT) ||
-            m_video_height == 0 || !Common::IsAlignedPow2(m_video_height, VIDEO_HEIGHT_ALIGNMENT)))
   {
     Error::SetStringView(error, "Invalid video dimensions/rate.");
     return false;