new

requirements.txt

@@ -0,0 +1,12 @@
+# Gradio UI
+gradio
+# Transformers + Diffusers
+transformers>=4.30.0
+diffusers>=0.16.0
+# PyTorch
+torch>=2.4.0
+# Wan-VAE and DiT dependencies
+xfuser>=0.4.1
+accelerate
+# Optional: huggingface_hub for model download
+huggingface_hub>=0.13.0

app.py

@@ -1,7 +1,42 @@
 import gradio as gr
+import torch
+from diffusers import DiffusionPipeline
+from huggingface_hub import hf_hub_download
 
-def greet(name):
-    return "Hello " + name + "!!"
+# Load model weights from Hub
+model_id = "Wan-AI/Wan2.1-I2V-14B-480P"
+ckpt_dir = hf_hub_download(repo_id=model_id, filename=".")
 
-demo = gr.Interface(fn=greet, inputs="text", outputs="text")
-demo.launch()
+pipe = DiffusionPipeline.from_pretrained(
+    model_id,
+    torch_dtype=torch.float16,
+    use_auth_token=True
+).to("cuda")
+
+pipe.enable_attention_slicing()
+
+
+def generate_video(image, prompt, num_frames=16):
+    video = pipe(
+        prompt=prompt,
+        init_image=image,
+        num_inference_steps=50,
+        guidance_scale=7.5,
+        num_frames=num_frames
+    ).videos
+    return video
+
+# Gradio UI
+def main():
+    with gr.Blocks() as demo:
+        gr.Markdown("# Wan2.1 Image-to-Video Demo")
+        with gr.Row():
+            img_in = gr.Image(type="pil", label="Input Image")
+            txt_p = gr.Textbox(label="Prompt")
+        btn = gr.Button("Generate Video")
+        out = gr.Video(label="Generated Video")
+        btn.click(fn=generate_video, inputs=[img_in, txt_p], outputs=out)
+    return demo
+
+if __name__ == "__main__":
+    main().launch()

spaces_config.json

@@ -0,0 +1,9 @@
+{
+    "sdk": "gradio",  
+    "hardware": {
+      "gpu": true,
+      "cpu": false,
+      "quantized": false
+    },
+    "timeout": 3600
+  }