Update app.py

app.py

@@ -185,12 +185,12 @@ elevenlabs_pipeline = ElevenLabsPipeline()
 stf_pipeline = STFPipeline()
 driving_video_path=gr.Video()
 
-@spaces.GPU(duration=120)
+@spaces.GPU(duration=240)
 def gpu_wrapped_stf_pipeline_execute(audio_path):
     return stf_pipeline.execute(audio_path)
 
 
-@spaces.GPU(duration=200)
+@spaces.GPU(duration=240)
 def gpu_wrapped_elevenlabs_pipeline_generate_voice(text, voice):
     return elevenlabs_pipeline.generate_voice(text, voice)
 
@@ -224,6 +224,7 @@ def txt_to_driving_video(text):
 # assets
 title_md = "assets/gradio_title.md"
 example_portrait_dir = "assets/examples/source"
+example_portrait_dir_custom = "assets/examples/source"
 example_video_dir = "assets/examples/driving"
 data_examples = [
     [osp.join(example_portrait_dir, "s9.jpg"), osp.join(example_video_dir, "d0.mp4"), True, True, True, True],
@@ -256,13 +257,17 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
             with gr.Row():
                 with gr.Column():
                     script_txt = gr.Text()
+                # with gr.Column():
+                #     txt2video_gen_button = gr.Button("txt2video generation", variant="primary")
+
                 with gr.Column():
-                    txt2video_gen_button = gr.Button("txt2video generation", variant="primary")
+                    audio_gen_button = gr.Button("Audio generation", variant="primary")
+            with gr.Row():
+                    output_audio = gr.Audio(label="Generated audio", type="filepath")
+            with gr.Row():
+                    video_gen_button = gr.Button("Audio to Video generation", variant="primary")
 
-                # with gr.Column():
-                #     audio_gen_button = gr.Button("Audio generation", variant="primary")
-            # with gr.Row():
-            #         video_input = gr.Audio(label="Generated video", type="filepath")
+                
                 
             gr.Markdown(load_description("assets/gradio_description_upload.md"))
             with gr.Row():
@@ -270,13 +275,14 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
                     image_input = gr.Image(type="filepath")
                     gr.Examples(
                         examples=[
-                            [osp.join(example_portrait_dir, "s9.jpg")],
-                            [osp.join(example_portrait_dir, "s6.jpg")],
-                            [osp.join(example_portrait_dir, "s10.jpg")],
-                            [osp.join(example_portrait_dir, "s5.jpg")],
-                            [osp.join(example_portrait_dir, "s7.jpg")],
-                            [osp.join(example_portrait_dir, "s12.jpg")],
-                            [osp.join(example_portrait_dir, "s22.jpg")],
+                            [osp.join(example_portrait_dir, "01.webp")],
+                            [osp.join(example_portrait_dir, "02.webp")],
+                            [osp.join(example_portrait_dir, "03.jpg")],
+                            [osp.join(example_portrait_dir, "04.jpg")],
+                            [osp.join(example_portrait_dir, "05.jpg")],
+                            [osp.join(example_portrait_dir, "06.jpg")],
+                            [osp.join(example_portrait_dir, "07.jpg")],
+                            [osp.join(example_portrait_dir, "08.jpg")],
                         ],
                         inputs=[image_input],
                         cache_examples=False,
@@ -345,11 +351,28 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
                 outputs=[output_video, output_video_concat],
                 show_progress=True
             )
-            txt2video_gen_button.click(
-                fn=txt_to_driving_video,
+            # txt2video_gen_button.click(
+            #     fn=txt_to_driving_video,
+            #     inputs=[
+            #         script_txt
+            #     ],
+            #     outputs=[video_input],
+            #     show_progress=True
+            # )
+            audio_gen_button.click(
+                fn=gpu_wrapped_elevenlabs_pipeline_generate_voice,
                 inputs=[
                     script_txt
                 ],
+                outputs=[output_audio],
+                show_progress=True
+            )
+
+            video_gen_button.click(
+                fn=gpu_wrapped_stf_pipeline_execute,
+                inputs=[
+                    output_audio
+                ],
                 outputs=[video_input],
                 show_progress=True
             )