Create app.py

app.py

@@ -0,0 +1,122 @@
+import gradio as gr
+import os
+import subprocess
+import sys
+import uuid
+from huggingface_hub import snapshot_download
+
+# --- 1. Environment Setup: Download the required base model ---
+# This runs only once when the Space starts.
+# The Ditto models are already in the repo, but the base model is separate.
+print("Downloading base model 'Wan-AI/Wan2.1-VACE-14B'...")
+try:
+    snapshot_download(
+        repo_id="Wan-AI/Wan2.1-VACE-14B",
+        local_dir="models/Wan-AI/Wan2.1-VACE-14B",
+        local_dir_use_symlinks=False # Use full downloads on Spaces
+    )
+    print("Base model downloaded successfully.")
+except Exception as e:
+    print(f"ERROR: Failed to download base model. The app may not work. Error: {e}")
+
+
+# --- 2. The Core Inference Function ---
+# This function wraps the command-line script provided in the model card.
+def run_video_edit(input_video_path, prompt_text):
+    if not input_video_path:
+        raise gr.Error("You must upload an input video.")
+    if not prompt_text or not prompt_text.strip():
+        raise gr.Error("You must provide an editing instruction.")
+
+    print(f"Starting video edit process for: {input_video_path}")
+    print(f"Instruction: {prompt_text}")
+
+    # Define paths for the script and the specific LoRA model to use
+    inference_script_path = "inference/infer_ditto.py"
+    lora_model_path = "models/lora/Editto-XL.safetensors" # Using the main XL model
+
+    # Create a unique path for the output video in a temporary directory
+    output_filename = f"{uuid.uuid4()}.mp4"
+    output_video_path = os.path.join("/tmp", output_filename)
+
+    # Construct the command as specified in the model card
+    command = [
+        sys.executable,  # Use the current python interpreter
+        inference_script_path,
+        "--input_video", input_video_path,
+        "--output_video", output_video_path,
+        "--prompt", prompt_text,
+        "--lora_path", lora_model_path,
+        "--num_frames", "73",  # Default from the model card's example
+        "--device_id", "0"
+    ]
+
+    print(f"Executing command: {' '.join(command)}")
+
+    # Run the subprocess and capture output for debugging
+    try:
+        process = subprocess.run(
+            command,
+            check=True,
+            capture_output=True,
+            text=True
+        )
+        print("Inference script stdout:")
+        print(process.stdout)
+        print("Inference script stderr:")
+        print(process.stderr)
+    except subprocess.CalledProcessError as e:
+        print("ERROR: The inference script failed.")
+        print("Return code:", e.returncode)
+        print("Stdout:", e.stdout)
+        print("Stderr:", e.stderr)
+        # Display the error to the user in the Gradio UI
+        raise gr.Error(f"The model script failed. Check the logs for details. Stderr: {e.stderr}")
+
+    if not os.path.exists(output_video_path):
+        raise gr.Error("Inference completed, but the output video file was not created. Check the logs.")
+
+    print(f"Process finished successfully. Output video at: {output_video_path}")
+    return output_video_path
+
+
+# --- 3. Build the Gradio User Interface ---
+with gr.Blocks(css="#col-container {max-width: 780px; margin: auto;}") as demo:
+    with gr.Column(elem_id="col-container"):
+        gr.Markdown(
+        """
+        # Ditto / Editto: Instruction-Based Video Editing
+        This demo uses the official inference script from the [QingyanBai/Ditto_models](https://huggingface.co/QingyanBai/Ditto_models) repository to edit videos.
+        Upload a video, provide a text instruction, and click "Edit Video".
+        **Note:** Running on a ZeroGPU, the first startup and each video process will take some time. Please be patient.
+        """
+        )
+
+        with gr.Row():
+            with gr.Column():
+                input_video = gr.Video(label="Input Video")
+                instruction = gr.Textbox(label="Editing Instruction", placeholder="e.g., make it snowing")
+                submit_btn = gr.Button("Edit Video", variant="primary")
+            with gr.Column():
+                output_video = gr.Video(label="Edited Video", interactive=False)
+
+        gr.Markdown("## Example Instructions")
+        gr.Examples(
+            examples=[
+                ["change the background to a beach"],
+                ["make it a cartoon"],
+                ["add fireworks to the sky"],
+                ["make it night"],
+                ["turn it into a watercolor painting"]
+            ],
+            inputs=[instruction],
+            label="Click an example to use it (you still need to upload a video)"
+        )
+
+    submit_btn.click(
+        fn=run_video_edit,
+        inputs=[input_video, instruction],
+        outputs=[output_video]
+    )
+
+demo.launch()