Bugfix: Layer Mask crashing without valid image

Removed unnecessary activate live-cam option
Added time info to preview frame and changed frame slider event to allow faster changes
Upgraded to gradio 4.21.0

README.md

@@ -58,6 +58,15 @@ Additional commandline arguments are currently unsupported and settings should b
 
 
 ### Changelog
+**18.03.2024** v3.6.0
+
+- Updated to Gradio 4.21.0 - requiring many changes under the hood
+- New manual masking (draw the mask yourself)
+- Extras Tab, streamlined cutting/joining videos
+- Re-added face selection by gender (on-demand loading, default turned off)
+- Removed unnecessary activate live-cam option
+- Added time info to preview frame and changed frame slider event to allow faster changes
+
 
 **10.03.2024** v3.5.5
 

config_colab.yaml

@@ -1,6 +1,5 @@
 clear_output: true
 force_cpu: false
-live_cam_start_active: false
 max_threads: 3
 memory_limit: 0
 output_image_format: png

requirements.txt

@@ -1,7 +1,7 @@
 --extra-index-url https://download.pytorch.org/whl/cu118
 
 numpy==1.26.2
-gradio==4.19.2
+gradio==4.21.0
 opencv-python==4.8.1.78
 onnx==1.15.0
 insightface==0.7.3
@@ -13,7 +13,7 @@ torchvision==0.16.2; sys_platform == 'darwin'
 onnxruntime==1.16.3; sys_platform == 'darwin' and platform_machine != 'arm64'
 onnxruntime-silicon==1.13.1; sys_platform == 'darwin' and platform_machine == 'arm64'
 onnxruntime-gpu==1.16.3; sys_platform != 'darwin'
-tqdm==4.66.1
+tqdm==4.66.2
 ftfy
 regex
 pyvirtualcam

settings.py

@@ -30,7 +30,6 @@ def load(self):
         self.output_video_codec = self.default_get(data, 'output_video_codec', 'libx264')
         self.video_quality = self.default_get(data, 'video_quality', 14)
         self.clear_output = self.default_get(data, 'clear_output', True)
-        self.live_cam_start_active = self.default_get(data, 'live_cam_start_active', False)
         self.max_threads = self.default_get(data, 'max_threads', 2)
         self.memory_limit = self.default_get(data, 'memory_limit', 0)
         self.provider = self.default_get(data, 'provider', 'cuda')
@@ -53,7 +52,6 @@ def save(self):
             'output_video_codec' : self.output_video_codec,
             'video_quality' : self.video_quality,
             'clear_output' : self.clear_output,
-            'live_cam_start_active' : self.live_cam_start_active,
             'max_threads' : self.max_threads,
             'memory_limit' : self.memory_limit,
             'provider' : self.provider,

ui/globals.py

@@ -7,8 +7,6 @@
 ui_blend_ratio = None
 ui_input_thumbs = []
 ui_target_thumbs = []
-
-ui_live_cam_active = False
 ui_camera_frame = None
 
 

ui/main.py

@@ -45,7 +45,6 @@ def run():
             overflow-y: auto !important;
         }
 """
-    uii.ui_live_cam_active = roop.globals.CFG.live_cam_start_active
 
     while run_server:
         server_name = roop.globals.CFG.server_name

ui/tabs/faceswap_tab.py

@@ -66,7 +66,7 @@ def faceswap_tab():
 
             with gr.Column(scale=2):
                 previewimage = gr.Image(label="Preview Image", height=576, interactive=False, visible=True)
-                maskimage = gr.ImageEditor(label="Manual mask Image", sources=["clipboard"], transforms="",
+                maskimage = gr.ImageEditor(label="Manual mask Image", sources=["clipboard"], transforms="", type="numpy",
                                              brush=gr.Brush(color_mode="fixed", colors=["rgba(255, 255, 255, 1"]), interactive=True, visible=False)
                 with gr.Row(variant='panel'):
                         fake_preview = gr.Checkbox(label="Face swap frames", value=False)
@@ -126,8 +126,8 @@ def faceswap_tab():
                 resultvideo = gr.Video(label='Final Video', interactive=False, visible=False)
 
     previewinputs = [preview_frame_num, bt_destfiles, fake_preview, ui.globals.ui_selected_enhancer, selected_face_detection,
-                        max_face_distance, ui.globals.ui_blend_ratio, chk_useclip, clip_text, no_face_action, vr_mode, autorotate, previewimage, maskimage]
-    previewoutputs = [preview_frame_num,previewimage, maskimage, mask_top, mask_bottom, mask_left, mask_right] 
+                        max_face_distance, ui.globals.ui_blend_ratio, chk_useclip, clip_text, no_face_action, vr_mode, autorotate, maskimage]
+    previewoutputs = [previewimage, maskimage, preview_frame_num] 
     input_faces.select(on_select_input_face, None, None).then(fn=on_preview_frame_changed, inputs=previewinputs, outputs=previewoutputs)
     bt_remove_selected_input_face.click(fn=remove_selected_input_face, outputs=[input_faces])
     bt_srcfiles.change(fn=on_srcfile_changed, show_progress='full', inputs=bt_srcfiles, outputs=[dynamic_face_selection, face_selection, input_faces])
@@ -168,7 +168,7 @@ def faceswap_tab():
     bt_refresh_preview.click(fn=on_preview_frame_changed, inputs=previewinputs, outputs=previewoutputs)            
     bt_toggle_masking.click(fn=on_toggle_masking, inputs=[previewimage, maskimage], outputs=[previewimage, maskimage])            
     fake_preview.change(fn=on_preview_frame_changed, inputs=previewinputs, outputs=previewoutputs)
-    preview_frame_num.change(fn=on_preview_frame_changed, inputs=previewinputs, outputs=previewoutputs, show_progress='hidden')
+    preview_frame_num.release(fn=on_preview_frame_changed, inputs=previewinputs, outputs=previewoutputs, show_progress='hidden', )
     bt_use_face_from_preview.click(fn=on_use_face_from_selected, show_progress='full', inputs=[bt_destfiles, preview_frame_num], outputs=[dynamic_face_selection, face_selection, target_faces, selected_face_detection])
     set_frame_start.click(fn=on_set_frame, inputs=[set_frame_start, preview_frame_num], outputs=[text_frame_clip])
     set_frame_end.click(fn=on_set_frame, inputs=[set_frame_end, preview_frame_num], outputs=[text_frame_clip])
@@ -373,8 +373,8 @@ def on_end_face_selection():
 
 
 def on_preview_frame_changed(frame_num, files, fake_preview, enhancer, detection, face_distance, blend_ratio,
-                              use_clip, clip_text, no_face_action, vr_mode, auto_rotate, previewimage, maskimage):
-    global SELECTED_INPUT_FACE_INDEX, is_processing, manual_masking
+                              use_clip, clip_text, no_face_action, vr_mode, auto_rotate, maskimage):
+    global SELECTED_INPUT_FACE_INDEX, manual_masking, current_video_fps
 
     from roop.core import live_swap
 
@@ -386,29 +386,33 @@ def on_preview_frame_changed(frame_num, files, fake_preview, enhancer, detection
         mask_offsets = roop.globals.INPUT_FACESETS[SELECTED_INPUT_FACE_INDEX].faces[0].mask_offsets
 
     timeinfo = '0:00:00'
-    if is_processing or files is None or selected_preview_index >= len(files) or frame_num is None:
-        return gr.Slider(info=timeinfo),None,gr.ImageEditor(visible=False), mask_offsets[0], mask_offsets[1], mask_offsets[2], mask_offsets[3]
+    if files is None or selected_preview_index >= len(files) or frame_num is None:
+        return None,None, gr.Slider(info=timeinfo)
 
     filename = files[selected_preview_index].name
     # time.sleep(0.3)
     if util.is_video(filename) or filename.lower().endswith('gif'):
         current_frame = get_video_frame(filename, frame_num)
-        secs = frame_num / current_video_fps
+        if current_video_fps == 0:
+            current_video_fps = 1
+        secs = (frame_num - 1) / current_video_fps
         minutes = secs / 60
         secs = secs % 60
         hours = minutes / 60
-        timeinfo = f"{int(hours):0>2}:{int(minutes):0>2}:{int(secs):0>2}"  
+        minutes = minutes % 60
+        milliseconds = (secs - int(secs)) * 1000
+        timeinfo = f"{int(hours):0>2}:{int(minutes):0>2}:{int(secs):0>2}.{int(milliseconds):0>3}"  
     else:
         current_frame = get_image_frame(filename)
     if current_frame is None:
-        return gr.Slider(info=timeinfo), gr.Image(visible=True), gr.ImageEditor(visible=False), mask_offsets[0], mask_offsets[1], mask_offsets[2], mask_offsets[3]
+        return None, None, gr.Slider(info=timeinfo)
 
     layers = None
     if maskimage is not None:
         layers = maskimage["layers"]
 
     if not fake_preview or len(roop.globals.INPUT_FACESETS) < 1:
-        return gr.Slider(info=timeinfo), gr.Image(value=util.convert_to_gradio(current_frame), visible=True), gr.ImageEditor(visible=False), mask_offsets[0], mask_offsets[1], mask_offsets[2], mask_offsets[3]
+        return gr.Image(value=util.convert_to_gradio(current_frame), visible=True), gr.ImageEditor(visible=False), gr.Slider(info=timeinfo)
 
     roop.globals.face_swap_mode = translate_swap_mode(detection)
     roop.globals.selected_enhancer = enhancer
@@ -425,8 +429,8 @@ def on_preview_frame_changed(frame_num, files, fake_preview, enhancer, detection
     mask = layers[0] if layers is not None else None
     current_frame = live_swap(current_frame, roop.globals.face_swap_mode, use_clip, clip_text, maskimage, SELECTED_INPUT_FACE_INDEX)
     if current_frame is None:
-        return gr.Slider(info=timeinfo), gr.Image(visible=True), gr.ImageEditor(visible=False), mask_offsets[0], mask_offsets[1], mask_offsets[2], mask_offsets[3] 
-    return gr.Slider(info=timeinfo), gr.Image(value=util.convert_to_gradio(current_frame), visible=True), gr.ImageEditor(visible=False), mask_offsets[0], mask_offsets[1], mask_offsets[2], mask_offsets[3]
+        return gr.Image(visible=True), None, gr.Slider(info=timeinfo)
+    return gr.Image(value=util.convert_to_gradio(current_frame), visible=True), gr.ImageEditor(visible=False), gr.Slider(info=timeinfo)
 
 def on_toggle_masking(previewimage, mask):
     global manual_masking
@@ -464,7 +468,7 @@ def on_preview_mask(frame_num, files, clip_text):
     from roop.core import preview_mask
     global is_processing
 
-    if is_processing:
+    if is_processing or files is None or selected_preview_index >= len(files) or clip_text is None or frame_num is None:
         return None
 
     filename = files[selected_preview_index].name

ui/tabs/livecam_tab.py

@@ -37,14 +37,12 @@ def start_cam(stream_to_obs, cam, reso, enhancer, blend_ratio):
     from roop.virtualcam import start_virtual_cam
     from roop.utilities import convert_to_gradio
 
-    ui.globals.ui_live_cam_active = True
     start_virtual_cam(stream_to_obs, cam, reso)
     roop.globals.selected_enhancer = enhancer
     roop.globals.blend_ratio = blend_ratio
-
-    while ui.globals.ui_live_cam_active:
+    while True:
         yield gr.Button(interactive=False), gr.Button(interactive=True), convert_to_gradio(ui.globals.ui_camera_frame)
-    return gr.Button(interactive=True), gr.Button(interactive=False), convert_to_gradio(ui.globals.ui_camera_frame)
+        
 
 def stop_swap():
     from roop.virtualcam import stop_virtual_cam

ui/tabs/settings_tab.py

@@ -45,7 +45,6 @@ def settings_tab():
             with gr.Column():
                 button_apply_restart = gr.Button("Restart Server", variant='primary')
                 with gr.Group():
-                    settings_controls.append(gr.Checkbox(label='Start with active live cam', value=roop.globals.CFG.live_cam_start_active, elem_id='live_cam_start_active', interactive=True))
                     settings_controls.append(gr.Checkbox(label='Use OS temp folder', value=roop.globals.CFG.use_os_temp_folder, elem_id='use_os_temp_folder', interactive=True))
                 button_clean_temp = gr.Button("Clean temp folder")
                 button_apply_settings = gr.Button("Apply Settings")