Update app.py

app.py

@@ -17,15 +17,15 @@ def load_audio(audio_listed):
 def execute_command(command: str) -> None:
     subprocess.run(command, check=True)
 
-def infer(audio_input, image_path):
+def infer(audio_input, image_path, emotional_style):
 
-    output_name = "acknowledgement_english@M030_front_neutral_level1_001@male_face"
+    output_name = "lipsynced_result"
 
     command = [
         f"python",
         f"inference_for_demo_video.py",
         f"--wav_path={audio_input}",
-        f"--style_clip_path=data/style_clip/3DMM/M030_front_neutral_level1_001.mat",
+        f"--style_clip_path=data/style_clip/3DMM/{emotional_style}",
         f"--pose_path=data/pose/RichardShelby_front_neutral_level1_001.mat",
         f"--image_path={image_path}",
         f"--cfg_scale=1.0",
@@ -46,8 +46,11 @@ def infer(audio_input, image_path):
 with gr.Blocks() as demo:
     with gr.Column():
         gr.HTML("""
-        <h2 style="text-align: center;">DreamTalk</h2>
-        <p style="text-align: center;"></p>
+        <h2 style="text-align: center;">DreamTalk: When Expressive Talking Head Generation Meets Diffusion Probabilistic Models</h2>
+        <p style="text-align: center;">
+            DreamTalk is a diffusion-based audio-driven expressive talking head generation framework that can produce high-quality talking head videos across diverse speaking styles. DreamTalk exhibits robust performance with a diverse array of inputs, including songs, speech in multiple languages, noisy audio, and out-of-domain portraits.
+            <img src="https://github.com/ali-vilab/dreamtalk/raw/main/media/teaser.gif" />
+        </p>
         """)
         with gr.Row():
             with gr.Column():
@@ -56,38 +59,77 @@ with gr.Blocks() as demo:
                     examples = [
                         "data/src_img/uncropped/face3.png",
                         "data/src_img/uncropped/male_face.png",
-                        "data/src_img/uncropped/uncut_src_img.jpg"
+                        "data/src_img/uncropped/uncut_src_img.jpg",
+                        "data/src_img/cropped/chpa5.png",
+                        "data/src_img/cropped/cut_img.png",
+                        "data/src_img/cropped/f30.png",
+                        "data/src_img/cropped/menglu2.png",
+                        "data/src_img/cropped/nscu2.png",
+                        "data/src_img/cropped/zp1.png"
+                        "data/src_img/cropped/zt12.png"
                     ],
                     inputs=[image_path]
                 )
-                audio_input = gr.Audio(label="Audio input", type="filepath", sources=["upload"])
-                audio_list = gr.Dropdown(
-                    label="Choose an audio (optional)",
-                    choices=[
-                        "German1.wav", "German2.wav", "German3.wav", "German4.wav",
-                        "acknowledgement_chinese.m4a", "acknowledgement_english.m4a",
-                        "chinese1_haierlizhi.wav", "chinese2_guanyu.wav",
-                        "french1.wav", "french2.wav", "french3.wav",
-                        "italian1.wav", "italian2.wav", "italian3.wav",
-                        "japan1.wav", "japan2.wav", "japan3.wav",
-                        "korean1.wav", "korean2.wav", "korean3.wav",
-                        "noisy_audio_cafeter_snr_0.wav", "noisy_audio_meeting_snr_0.wav", "noisy_audio_meeting_snr_10.wav", "noisy_audio_meeting_snr_20.wav", "noisy_audio_narrative.wav", "noisy_audio_office_snr_0.wav", "out_of_domain_narrative.wav",
-                        "spanish1.wav", "spanish2.wav", "spanish3.wav"
-                    ]
-                )
-                audio_list.change(
-                    fn = load_audio,
-                    inputs = [audio_list],
-                    outputs = [audio_input]
-                )
+                audio_input = gr.Audio(label="Audio input", type="filepath", sources=["upload"], value="data/audio/acknowledgement_english.m4a")
+                with gr.Row():
+                    audio_list = gr.Dropdown(
+                        label="Choose an audio (optional)",
+                        choices=[
+                            "German1.wav", "German2.wav", "German3.wav", "German4.wav",
+                            "acknowledgement_chinese.m4a", "acknowledgement_english.m4a",
+                            "chinese1_haierlizhi.wav", "chinese2_guanyu.wav",
+                            "french1.wav", "french2.wav", "french3.wav",
+                            "italian1.wav", "italian2.wav", "italian3.wav",
+                            "japan1.wav", "japan2.wav", "japan3.wav",
+                            "korean1.wav", "korean2.wav", "korean3.wav",
+                            "noisy_audio_cafeter_snr_0.wav", "noisy_audio_meeting_snr_0.wav", "noisy_audio_meeting_snr_10.wav", "noisy_audio_meeting_snr_20.wav", "noisy_audio_narrative.wav", "noisy_audio_office_snr_0.wav", "out_of_domain_narrative.wav",
+                            "spanish1.wav", "spanish2.wav", "spanish3.wav"
+                        ],
+                        value = "acknowledgement_english.m4a"
+                    )
+                    audio_list.change(
+                        fn = load_audio,
+                        inputs = [audio_list],
+                        outputs = [audio_input]
+                    )
+                    emotional_style = gr.Dropdown(
+                        label = "emotional style",
+                        choices = [
+                            "M030_front_angry_level3_001.mat",
+                            "M030_front_contempt_level3_001.mat",
+                            "M030_front_disgusted_level3_001.mat",
+                            "M030_front_fear_level3_001.mat",
+                            "M030_front_happy_level3_001.mat",
+                            "M030_front_neutral_level1_001.mat",
+                            "M030_front_sad_level3_001.mat",
+                            "M030_front_surprised_level3_001.mat",
+                            "W009_front_angry_level3_001.mat",
+                            "W009_front_contempt_level3_001.mat",
+                            "W009_front_disgusted_level3_001.mat",
+                            "W009_front_fear_level3_001.mat",
+                            "W009_front_happy_level3_001.mat",
+                            "W009_front_neutral_level1_001.mat",
+                            "W009_front_sad_level3_001.mat",
+                            "W009_front_surprised_level3_001.mat",
+                            "W011_front_angry_level3_001.mat",
+                            "W011_front_contempt_level3_001.mat",
+                            "W011_front_disgusted_level3_001.mat",
+                            "W011_front_fear_level3_001.mat",
+                            "W011_front_happy_level3_001.mat",
+                            "W011_front_neutral_level1_001.mat",
+                            "W011_front_sad_level3_001.mat",
+                            "W011_front_surprised_level3_001.mat
+                        ],
+                        value = "M030_front_neutral_level1_001.mat"
+                    )
                 run_btn = gr.Button("Run")
             with gr.Column():
                 output_video = gr.Video(format="mp4")
     
     run_btn.click(
         fn = infer,
-        inputs = [audio_input, image_path],
+        inputs = [audio_input, image_path, emotional_style],
         outputs = [output_video]
     )
 
-demo.launch()
+demo.queue().launch()