app.py

import os
import gradio as gr
import torch
import subprocess
from PIL import Image
from pathlib import Path
import io
import sys
import traceback

# =========================================
# 1. Define Hugging Face weights and paths
# =========================================
from huggingface_hub import hf_hub_download

HF_DATASET_REPO = "roll-ai/FloVD-weights"  # dataset repo ID

def download_weights():
    print("🔄 Downloading model weights via huggingface_hub...")
    for rel_path in WEIGHT_FILES.values():
        local_path = Path("ckpt") / rel_path
        if not local_path.exists():
            print(f"📥 Downloading {rel_path}")
            hf_hub_download(
                repo_id=HF_DATASET_REPO,
                repo_type="dataset",
                filename=rel_path,
                local_dir="ckpt",
                local_dir_use_symlinks=False,
            )
        else:
            print(f"✅ Already exists: {local_path}")

from inference.flovd_demo import generate_video

def run_inference(prompt, image, pose_type, speed, use_flow_integration, cam_pose_name):
    # Redirect stdout to capture logs
    log_buffer = io.StringIO()
    sys_stdout = sys.stdout
    sys.stdout = log_buffer

    video_path = None
    try:
        print("🚀 Starting inference...")
        os.makedirs("input_images", exist_ok=True)
        image_path = "input_images/input_image.png"
        image.save(image_path)
        print(f"📸 Saved input image to {image_path}")

        generate_video(
            prompt=prompt,
            image_path=image_path,
            fvsm_path="./ckpt/FVSM/FloVD_FVSM_Controlnet.pt",
            omsm_path="./ckpt/OMSM",
            output_path="./outputs",
            num_frames=49,
            fps=16,
            width=None,
            height=None,
            seed=42,
            guidance_scale=6.0,
            dtype=torch.float16,
            controlnet_guidance_end=0.4,
            use_dynamic_cfg=False,
            pose_type=pose_type,
            speed=float(speed),
            use_flow_integration=use_flow_integration,
            cam_pose_name=cam_pose_name,
            depth_ckpt_path="./ckpt/others/depth_anything_v2_metric_hypersim_vitb.pth"
        )

        video_name = f"{prompt[:30].strip().replace(' ', '_')}_{cam_pose_name or 'default'}.mp4"
        video_path = f"./outputs/generated_videos/{video_name}"
        print(f"✅ Inference complete. Video saved to {video_path}")

    except Exception as e:
        print("🔥 Inference failed with exception:")
        traceback.print_exc()

    # Restore stdout and return logs
    sys.stdout = sys_stdout
    logs = log_buffer.getvalue()
    log_buffer.close()

    return (video_path if video_path and os.path.exists(video_path) else None), logs

# ========================
# Gradio Interface
# ========================

with gr.Blocks() as demo:
    gr.Markdown("## 🎥 FloVD: Optical Flow + CogVideoX Video Generation")
    with gr.Row():
        with gr.Column():
            prompt = gr.Textbox(label="Prompt", value="A girl riding a bicycle through a park.")
            image = gr.Image(type="pil", label="Input Image")
            pose_type = gr.Radio(choices=["manual", "re10k"], value="manual", label="Camera Pose Type")
            cam_pose_name = gr.Textbox(label="Camera Trajectory Name", placeholder="e.g. zoom_in, tilt_up")
            speed = gr.Slider(minimum=0.1, maximum=2.0, step=0.1, value=0.5, label="Speed")
            use_flow_integration = gr.Checkbox(label="Use Flow Integration", value=False)
            submit = gr.Button("Generate Video")
        with gr.Column():
            output_video = gr.Video(label="Generated Video")
            output_logs = gr.Textbox(label="Logs", lines=20, interactive=False)

    submit.click(
        fn=run_inference,
        inputs=[prompt, image, pose_type, speed, use_flow_integration, cam_pose_name],
        outputs=[output_video, output_logs]
    )
demo.launch(show_error=True)