Create app.py

app.py

@@ -0,0 +1,54 @@
+import os
+import time
+import gradio as gr
+import torch
+from diffusers import DiffusionPipeline
+from PIL import Image
+import numpy as np
+import cv2
+from moviepy.editor import ImageSequenceClip
+
+class WanAnimateApp:
+    def __init__(self):
+        model_name = "Wan-AI/Wan2.2-Animate-14B"
+        self.pipe = DiffusionPipeline.from_pretrained(
+            model_name,
+            torch_dtype=torch.float32,
+            use_safetensors=True,
+            device_map="cpu"
+        )
+        self.pipe.reset_device_map()
+        self.pipe.enable_model_cpu_offload()
+        self.pipe.enable_vae_slicing()
+
+    def predict(self, ref_img, video, model_id, model):
+        if ref_img is None:
+            return None, "Upload image."
+        try:
+            ref_image = Image.open(ref_img).convert("RGB").resize((512, 512))
+            motion = " with motion transfer" if video else ""
+            prompt = f"Animate character from image{motion}, detailed, high quality."
+            num_steps = 15 if model == "wan-pro" else 10
+            output = self.pipe(prompt, image=ref_image, num_inference_steps=num_steps, height=512, width=512).frames
+            temp_video = f"/tmp/out_{int(time.time())}.mp4"
+            clip = ImageSequenceClip(output, fps=10)
+            clip.write_videofile(temp_video, fps=10, codec='libx264', audio=False)
+            return temp_video, "Done!"
+        except Exception as e:
+            return None, str(e)
+
+app = WanAnimateApp()
+
+with gr.Blocks(title="Wan2.2-Animate CPU") as demo:
+    gr.HTML("<h1>Wan2.2-Animate (CPU Mode)</h1><p>Upload image for animation. Slow on CPU (10–30 min).</p>")
+    with gr.Row():
+        ref_img = gr.Image(label="Image", type="filepath")
+        video = gr.Video(label="Video (optional)")
+        model_id = gr.Dropdown(["move", "mix"], value="move", label="Mode")
+        model = gr.Dropdown(["wan-pro", "wan-std"], value="wan-std", label="Quality")
+        btn = gr.Button("Generate")
+    output_video = gr.Video(label="Output")
+    status = gr.Textbox(label="Status")
+    btn.click(app.predict, inputs=[ref_img, video, model_id, model], outputs=[output_video, status])
+
+demo.queue().launch(server_name="0.0.0.0", server_port=7860)