Update app.py

app.py

@@ -516,7 +516,7 @@ character_name_to_yaml = {
 }
 
 @spaces.GPU(duration=200) 
-def tango(audio_path, character_name, seed, create_graph=False, video_folder_path=None):
+def tango(audio_path, character_name, seed, threshold=1.0, create_graph=False, video_folder_path=None):
     os.system("rm -r ./outputs/")
     cfg = prepare_all("./configs/gradio.yaml")
     cfg.seed = seed
@@ -551,7 +551,7 @@ def tango(audio_path, character_name, seed, create_graph=False, video_folder_pat
         graph_save_path = "./outputs/save_video.pkl"
         os.system(f"cd ./SMPLer-X/ && python app.py --video_folder_path .{video_folder_path} --data_save_path .{data_save_path} --json_save_path .{json_save_path} && cd ..")
         print(f"cd ./SMPLer-X/ && python app.py --video_folder_path .{video_folder_path} --data_save_path .{data_save_path} --json_save_path .{json_save_path} && cd ..")
-        os.system(f"python ./create_graph.py --json_save_path {json_save_path} --graph_save_path {graph_save_path}") 
+        os.system(f"python ./create_graph.py --json_save_path {json_save_path} --graph_save_path {graph_save_path} --threshold {threshold}") 
         cfg.data.test_meta_paths = json_save_path
         gc.collect()
         torch.cuda.empty_cache()
@@ -609,7 +609,7 @@ examples_video = [
 ]
 
 combined_examples = [
-    ["./datasets/cached_audio/example_female_voice_9_seconds.wav", "./datasets/cached_audio/female_test_V1.mp4", 2024],
+    ["./datasets/cached_audio/example_female_voice_9_seconds.wav", "./datasets/cached_audio/female_test_V1.mp4", 2024, 1.0],
     # ["./datasets/cached_audio/example_female_voice_9_seconds.wav", "./datasets/cached_audio/101099-00_18_09-00_18_19.mp4", 2024],
 ]
 
@@ -664,10 +664,12 @@ def make_demo():
         with gr.Row():
             gr.Markdown(
               """
-              <div style="display: flex; justify-content: center; align-items: center; text-align: center;">
+              <div style="display: flex; justify-content: center; align-items: center; text-align: left;">
+              News:
+              <br>
               This is an open-source project supported by Hugging Face's free L40S GPU. Runtime is limited, so it operates in low-quality mode. Some generated results from high-quality mode are shown above.
               <br>
-              News:
+              [10/19]: Add a parameter to control "smooth" or "jumpy" for custom uploaded video. lower threshold is more smooth but more repeated gestures.
               <br>
               [10/15]: Add watermark, fix bugs on custom character by downgrades to py3.9, fix bugs to support audio less than 4s.
               <br>
@@ -716,7 +718,7 @@ def make_demo():
         with gr.Row():
             with gr.Column(scale=1):
                 audio_input = gr.Audio(label="Upload your audio")
-                seed_input = gr.Number(label="Seed", value=2024, interactive=True)
+                seed_input = gr.Number(label="Random Seed", value=2024, interactive=True)
             with gr.Column(scale=2):
                 gr.Examples(
                     examples=examples_audio,
@@ -727,6 +729,7 @@ def make_demo():
                 )
             with gr.Column(scale=1):
                 video_input = gr.Video(label="Your Character", elem_classes="video")
+                # threshold_input = gr.Slider(label="Result's Smoothness, lower is more smooth, only effective for custom uploaded videos.", minimum=0.7, maximum=1.0, value=1.0, step=0.01, interactive=True)
             with gr.Column(scale=2):
                 gr.Examples(
                     examples=examples_video,
@@ -735,6 +738,7 @@ def make_demo():
                     label="Character Examples",
                     cache_examples=False
                 )
+                threshold_input = gr.Slider(label="Result's Smoothness, lower is more smooth, only effective for custom uploaded videos.", minimum=0.7, maximum=1.0, value=1.0, step=0.01, interactive=True)
         
         # Fourth row: Generate video button
         with gr.Row():
@@ -743,7 +747,7 @@ def make_demo():
         # Define button click behavior
         run_button.click(
             fn=tango,
-            inputs=[audio_input, video_input, seed_input],
+            inputs=[audio_input, video_input, seed_input, threshold_input],
             outputs=[video_output_1, video_output_2, file_output_1, file_output_2]
         )
 
@@ -751,7 +755,7 @@ def make_demo():
             with gr.Column(scale=4):
                 gr.Examples(
                     examples=combined_examples,
-                    inputs=[audio_input, video_input, seed_input],  # Both audio and video as inputs
+                    inputs=[audio_input, video_input, seed_input, threshold_input],  # Both audio and video as inputs
                     outputs=[video_output_1, video_output_2, file_output_1, file_output_2],
                     fn=tango,  # Function that processes both audio and video inputs
                     label="Select Combined Audio and Video Examples (Cached)",