Start with reference component

roop/uis/__components__/preview.py

@@ -29,12 +29,12 @@ def render() -> None:
         if is_image(roop.globals.target_path):
             temp_frame = cv2.imread(roop.globals.target_path)
             preview_frame = get_preview_frame(temp_frame)
-            preview_image_args['value'] = normalize_preview_frame(preview_frame)
+            preview_image_args['value'] = ui.normalize_frame(preview_frame)
             preview_image_args['visible'] = True
         if is_video(roop.globals.target_path):
             temp_frame = get_video_frame(roop.globals.target_path, roop.globals.reference_frame_number)
             preview_frame = get_preview_frame(temp_frame)
-            preview_image_args['value'] = normalize_preview_frame(preview_frame)
+            preview_image_args['value'] = ui.normalize_frame(preview_frame)
             preview_image_args['visible'] = True
             preview_slider_args['value'] = roop.globals.reference_frame_number
             preview_slider_args['maximum'] = get_video_frame_total(roop.globals.target_path)
@@ -54,12 +54,12 @@ def update(frame_number: int = 0) -> Tuple[Dict[Any, Any], Dict[Any, Any]]:
     if is_image(roop.globals.target_path):
         temp_frame = cv2.imread(roop.globals.target_path)
         preview_frame = get_preview_frame(temp_frame)
-        return gradio.update(value=normalize_preview_frame(preview_frame), visible=True), gradio.update(value=0, maximum=1, visible=False)
+        return gradio.update(value=ui.normalize_frame(preview_frame), visible=True), gradio.update(value=0, maximum=1, visible=False)
     if is_video(roop.globals.target_path):
         video_frame_total = get_video_frame_total(roop.globals.target_path)
         temp_frame = get_video_frame(roop.globals.target_path, frame_number)
         preview_frame = get_preview_frame(temp_frame)
-        return gradio.update(value=normalize_preview_frame(preview_frame), visible=True), gradio.update(value=frame_number, maximum=video_frame_total, visible=True)
+        return gradio.update(value=ui.normalize_frame(preview_frame), visible=True), gradio.update(value=frame_number, maximum=video_frame_total, visible=True)
     return gradio.update(value=None, visible=False), gradio.update(value=0, maximum=1, visible=False)
 
 
@@ -83,5 +83,3 @@ def get_preview_frame(temp_frame: Frame) -> Frame:
     return temp_frame
 
 
-def normalize_preview_frame(preview_frame: Frame) -> Frame:
-    return cv2.cvtColor(preview_frame, cv2.COLOR_BGR2RGB)

roop/uis/__components__/reference.py

@@ -0,0 +1,42 @@
+from time import sleep
+from typing import Dict, Any
+
+import gradio
+
+import roop.globals
+from roop.capturer import get_video_frame
+from roop.face_analyser import get_many_faces
+from roop.uis import core as ui
+from roop.utilities import is_video
+
+
+def render() -> None:
+    with gradio.Box():
+        reference_face_gallery_args: Dict[str, Any] = {
+            'label': 'reference_faces',
+            'columns': 6,
+            'height': 200,
+            'allow_preview': False,
+            'visible': True
+        }
+        if is_video(roop.globals.target_path):
+            target_frame = get_video_frame(roop.globals.target_path, roop.globals.reference_frame_number)
+            faces = get_many_faces(target_frame)
+            if faces:
+                value = []
+                for face in faces:
+                    start_x, start_y, end_x, end_y = map(int, face['bbox'])
+                    crop_frame = target_frame[start_y:end_y, start_x:end_x]
+                    value.append(ui.normalize_frame(crop_frame))
+                reference_face_gallery_args['value'] = value
+            else:
+                reference_face_gallery_args['value'] = []
+        reference_face_gallery = gradio.Gallery(**reference_face_gallery_args)
+        similar_face_distance_slider = gradio.Slider(
+            label='similar_face_distance',
+            value=roop.globals.similar_face_distance,
+            minimum=0,
+            maximum=2,
+            step=0.05
+        )
+        ui.register_component('reference_face_gallery', reference_face_gallery)

roop/uis/__components__/settings.py

@@ -1,4 +1,4 @@
-from typing import Any, Dict, List, Literal
+from typing import Any, Dict, List
 import gradio
 import onnxruntime
 

roop/uis/__layouts__/default.py

@@ -1,6 +1,6 @@
 import gradio
 
-from roop.uis.__components__ import settings, source, target, preview, output
+from roop.uis.__components__ import settings, source, target, preview, reference, output
 
 
 def render() -> gradio.Blocks:
@@ -13,6 +13,7 @@ def render() -> gradio.Blocks:
                 target.render()
             with gradio.Column(scale=2):
                 preview.render()
+                reference.render()
         with gradio.Row():
             output.render()
     return layout

roop/uis/core.py

@@ -1,8 +1,11 @@
 import importlib
 from typing import Dict, Optional, List
+
+import cv2
 import gradio
 
 import roop.globals
+from roop.typing import Frame
 from roop.uis.typing import Component, ComponentName
 from roop.utilities import list_module_names
 
@@ -34,3 +37,7 @@ def register_component(name: ComponentName, component: Component) -> None:
 
 def list_ui_layouts_names() -> Optional[List[str]]:
     return list_module_names('roop/uis/__layouts__')
+
+
+def normalize_frame(frame: Frame) -> Frame:
+    return cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)

roop/uis/typing.py

@@ -5,6 +5,7 @@ Component = gradio.File or gradio.Image or gradio.Video
 ComponentName = Literal[
     'source_file',
     'target_file',
+    'reference_face_gallery',
     'frame_processors_checkbox_group',
     'many_faces_checkbox'
 ]