wav2lip_gan support

app.py

@@ -31,11 +31,9 @@ def download_model():
     REPO_ID = 'vinthony/SadTalker-V002rc'
     snapshot_download(repo_id=REPO_ID, local_dir='./checkpoints', local_dir_use_symlinks=True)
 
-    # Manually replace the Wav2Lip model with wav2lip_gan.pth
+    # Descargar wav2lip_gan directamente desde GitHub
     wav2lip_model_path = os.path.join('./checkpoints', 'wav2lip.pth')
-    if not os.path.exists(wav2lip_model_path):
-        os.system(
-            f"wget https://github.com/Rudrabha/Wav2Lip/releases/download/v1.0/wav2lip_gan.pth -O {wav2lip_model_path}")
+    os.system(f"wget https://github.com/Rudrabha/Wav2Lip/releases/download/v1.0/wav2lip_gan.pth -O {wav2lip_model_path}")
 
     print("Replaced Wav2Lip model with Wav2Lip GAN.")
 
@@ -225,7 +223,7 @@ def sadtalker_demo():
 if __name__ == "__main__":
 
     demo = sadtalker_demo()
-    demo.queue(max_size=10, api_open=True)
+    demo.queue(max_size=100, api_open=True)
     demo.launch(debug=True)
 
 

src/gradio_demo.py

@@ -164,7 +164,7 @@ class SadTalker():
             torch.cuda.synchronize()
             
         import gc; gc.collect()
-        
+
         return return_path
 
     

test_api.py

@@ -0,0 +1,60 @@
+import gradio as gr
+
+def call_sadtalker_api(source_image, driven_audio, preprocess_type, is_still_mode, enhancer, batch_size, size_of_image, pose_style, facerender, exp_weight, use_ref_video, ref_video, ref_info, use_idle_mode, length_of_audio, blink_every):
+    # Define the API endpoint
+    api_url = "http://localhost:7860/api/predict"
+
+    # Prepare the payload
+    payload = {
+        "source_image": source_image,
+        "driven_audio": driven_audio,
+        "preprocess_type": preprocess_type,
+        "is_still_mode": is_still_mode,
+        "enhancer": enhancer,
+        "batch_size": batch_size,
+        "size_of_image": size_of_image,
+        "pose_style": pose_style,
+        "facerender": facerender,
+        "exp_weight": exp_weight,
+        "use_ref_video": use_ref_video,
+        "ref_video": ref_video,
+        "ref_info": ref_info,
+        "use_idle_mode": use_idle_mode,
+        "length_of_audio": length_of_audio,
+        "blink_every": blink_every
+    }
+
+    # Make the API request
+    response = requests.post(api_url, json=payload)
+    result = response.json()
+
+    # Return the generated video URL
+    return result["data"]
+
+# Create the Gradio interface
+iface = gr.Interface(
+    fn=call_sadtalker_api,
+    inputs=[
+        gr.Image(type="filepath", label="Source Image"),
+        gr.Audio(type="filepath", label="Driven Audio"),
+        gr.Radio(["crop", "resize", "full", "extcrop", "extfull"], label="Preprocess Type"),
+        gr.Checkbox(label="Still Mode"),
+        gr.Checkbox(label="Enhancer"),
+        gr.Slider(minimum=1, maximum=10, step=1, label="Batch Size"),
+        gr.Radio([256, 512], label="Size of Image"),
+        gr.Slider(minimum=0, maximum=45, step=1, label="Pose Style"),
+        gr.Radio(["facevid2vid", "pirender"], label="Face Render"),
+        gr.Slider(minimum=0, maximum=3, step=0.1, label="Expression Weight"),
+        gr.Checkbox(label="Use Reference Video"),
+        gr.Video(label="Reference Video"),
+        gr.Radio(["pose", "blink", "pose+blink", "all"], label="Reference Info"),
+        gr.Checkbox(label="Use Idle Mode"),
+        gr.Number(label="Length of Audio"),
+        gr.Checkbox(label="Blink Every")
+    ],
+    outputs=gr.Video(label="Generated Video"),
+    title="SadTalker API Client"
+)
+
+# Launch the interface
+iface.launch()