Sticky preview only for landscape images and videos

facefusion/typing.py

@@ -68,6 +68,7 @@ MelFilterBank = NDArray[Any]
 
 Fps = float
 Padding = Tuple[int, int, int, int]
+Orientation = Literal['landscape', 'portrait']
 Resolution = Tuple[int, int]
 
 ProcessState = Literal['checking', 'processing', 'stopping', 'pending']

facefusion/uis/assets/overrides.css

@@ -87,3 +87,10 @@
 {
 	object-fit: cover;
 }
+
+:root:root:root:root .image-preview.is-landscape
+{
+	position: sticky;
+	top: 0;
+	z-index: 100;
+}

facefusion/uis/components/preview.py

@@ -17,7 +17,7 @@ from facefusion.processors.core import get_processors_modules
 from facefusion.typing import AudioFrame, Face, FaceSet, VisionFrame
 from facefusion.uis.core import get_ui_component, get_ui_components, register_ui_component
 from facefusion.uis.typing import ComponentOptions
-from facefusion.vision import count_video_frame_total, get_video_frame, normalize_frame_color, read_static_image, read_static_images, resize_frame_resolution
+from facefusion.vision import count_video_frame_total, detect_frame_orientation, get_video_frame, normalize_frame_color, read_static_image, read_static_images, resize_frame_resolution
 
 PREVIEW_IMAGE : Optional[gradio.Image] = None
 PREVIEW_FRAME_SLIDER : Optional[gradio.Slider] = None
@@ -56,11 +56,13 @@ def render() -> None:
 		target_vision_frame = read_static_image(state_manager.get_item('target_path'))
 		preview_vision_frame = process_preview_frame(reference_faces, source_face, source_audio_frame, target_vision_frame)
 		preview_image_options['value'] = normalize_frame_color(preview_vision_frame)
+		preview_image_options['elem_classes'] = [ 'image-preview', 'is-' + detect_frame_orientation(preview_vision_frame) ]
 
 	if is_video(state_manager.get_item('target_path')):
 		temp_vision_frame = get_video_frame(state_manager.get_item('target_path'), state_manager.get_item('reference_frame_number'))
 		preview_vision_frame = process_preview_frame(reference_faces, source_face, source_audio_frame, temp_vision_frame)
 		preview_image_options['value'] = normalize_frame_color(preview_vision_frame)
+		preview_image_options['elem_classes'] = [ 'image-preview', 'is-' + detect_frame_orientation(preview_vision_frame) ]
 		preview_image_options['visible'] = True
 		preview_frame_slider_options['value'] = state_manager.get_item('reference_frame_number')
 		preview_frame_slider_options['maximum'] = count_video_frame_total(state_manager.get_item('target_path'))
@@ -202,14 +204,14 @@ def update_preview_image(frame_number : int = 0) -> gradio.Image:
 		target_vision_frame = read_static_image(state_manager.get_item('target_path'))
 		preview_vision_frame = process_preview_frame(reference_faces, source_face, source_audio_frame, target_vision_frame)
 		preview_vision_frame = normalize_frame_color(preview_vision_frame)
-		return gradio.Image(value = preview_vision_frame)
+		return gradio.Image(value = preview_vision_frame, elem_classes = [ 'image-preview', 'is-' + detect_frame_orientation(preview_vision_frame) ])
 
 	if is_video(state_manager.get_item('target_path')):
 		temp_vision_frame = get_video_frame(state_manager.get_item('target_path'), frame_number)
 		preview_vision_frame = process_preview_frame(reference_faces, source_face, source_audio_frame, temp_vision_frame)
 		preview_vision_frame = normalize_frame_color(preview_vision_frame)
-		return gradio.Image(value = preview_vision_frame)
+		return gradio.Image(value = preview_vision_frame, elem_classes = [ 'image-preview', 'is-' + detect_frame_orientation(preview_vision_frame) ])
-	return gradio.Image(value = None)
+	return gradio.Image(value = None, elem_classes = None)
 
 
 def update_preview_frame_slider() -> gradio.Slider:

facefusion/vision.py

@@ -8,7 +8,7 @@ from cv2.typing import Size
 from facefusion.choices import image_template_sizes, video_template_sizes
 from facefusion.common_helper import is_windows
 from facefusion.filesystem import is_image, is_video, sanitize_path_for_windows
-from facefusion.typing import Fps, Resolution, VisionFrame
+from facefusion.typing import Fps, Orientation, Resolution, VisionFrame
 
 
 @lru_cache(maxsize = 128)
@@ -178,6 +178,14 @@ def unpack_resolution(resolution : str) -> Resolution:
 	return width, height
 
 
+def detect_frame_orientation(vision_frame : VisionFrame) -> Orientation:
+	height, width = vision_frame.shape[:2]
+
+	if width > height:
+		return 'landscape'
+	return 'portrait'
+
+
 def resize_frame_resolution(vision_frame : VisionFrame, max_resolution : Resolution) -> VisionFrame:
 	height, width = vision_frame.shape[:2]
 	max_width, max_height = max_resolution