update

app.py

@@ -1,20 +1,34 @@
 import os
+import subprocess
 import gdown
 import gradio as gr
 
-# Make checkpoints dir
-os.makedirs("checkpoints", exist_ok=True)
+# ---------------------------------------------------
+# AUTO CLONE SADTALKER FROM GITHUB (only first time)
+# ---------------------------------------------------
+
+if not os.path.exists("SadTalker"):
+    print("Cloning SadTalker from GitHub...")
+    subprocess.run(["git", "clone", "https://github.com/OpenTalker/SadTalker.git"])
+
+# Move inside SadTalker folder
+os.chdir("SadTalker")
+
+# ---------------------------------------------------
+# GOOGLE DRIVE DOWNLOADER
+# ---------------------------------------------------
 
-# Google Drive Download Helper
 def gd(file_id, out):
     url = f"https://drive.google.com/uc?id={file_id}"
     if not os.path.exists(out):
-        print(f"Downloading {out}...")
+        print(f"Downloading {out} ...")
         gdown.download(url, out, quiet=False)
     else:
         print(f"{out} already exists.")
 
-# ---- DOWNLOAD ALL MODELS ----
+os.makedirs("checkpoints", exist_ok=True)
+
+# Download all required model files
 gd("1g3VIpU3yhpITMZtrWU2mbmyzLoF9K3Gz", "checkpoints/auido2exp_00300-model.pth")
 gd("1Jp4i_Qc-6qCms7v1kN61RE3qnT_9J8vg", "checkpoints/auido2pose_00140-model.pth")
 gd("1pTbKmpOeWRSA1NYQ3DnWe3W-9Qeyy7PK", "checkpoints/mapping_00109-model.pth.tar")
@@ -23,10 +37,51 @@ gd("1caz3ESiy-aEzI0ptff8qpjcxdTf-L26E", "checkpoints/SadTalker_256.safetensors")
 gd("12mj9EEs4KIS6-9FjF-PjEegF-TDKvvi6", "checkpoints/SadTalker_512.safetensors")
 gd("1cag0u7e5RgdKoxBa7exY599VNGtYgNQb", "checkpoints/shape_predictor_68_face_landmarks.dat")
 
-# -------------------------------
+# ---------------------------------------------------
+# SADTALKER INFERENCE FUNCTION
+# ---------------------------------------------------
+
+def generate_video(image, audio):
+    # remove previous output
+    if os.path.exists("result.mp4"):
+        os.remove("result.mp4")
+
+    # Save input image & audio
+    image.save("input.png")
+    audio_path = "input.wav"
+    audio.save(audio_path)
+
+    # Run SadTalker inference
+    cmd = [
+        "python3", "inference.py",
+        "--driven_audio", audio_path,
+        "--source_image", "input.png",
+        "--result_dir", "./",
+        "--enhancer", "gfpgan",
+        "--still",  # input stability
+    ]
+
+    subprocess.run(cmd)
+
+    # Return output
+    if os.path.exists("result.mp4"):
+        return "result.mp4"
+    else:
+        return "Error: video not generated"
+
+# ---------------------------------------------------
+# GRADIO UI
+# ---------------------------------------------------
 
-def greet(name):
-    return "Hello " + name + "!!"
+demo = gr.Interface(
+    fn=generate_video,
+    inputs=[
+        gr.Image(type="pil", label="Source Image"),
+        gr.Audio(type="filepath", label="Input Voice")
+    ],
+    outputs=gr.Video(label="Generated Talking Video"),
+    title="SadTalker Online API (Auto-Clone + Google Drive Models)",
+    description="Upload image + audio → Generate talking video"
+)
 
-demo = gr.Interface(fn=greet, inputs="text", outputs="text")
 demo.launch()