app.py

#!/usr/bin/env python
"""
Gradio demo for Wan2.1 First-Last-Frame-to-Video (FLF2V)
Loads the huge model once, uses balanced device placement,
streams high-level progress, and auto-offers the .mp4 for download.
"""
import os
import numpy as np
import torch
import gradio as gr
from diffusers import WanImageToVideoPipeline, AutoencoderKLWan
from diffusers.utils import export_to_video
from transformers import CLIPImageProcessor, CLIPVisionModel
from PIL import Image
import torchvision.transforms.functional as TF

# --------------------------------------------------------------------
# CONFIG
MODEL_ID     = "Wan-AI/Wan2.1-FLF2V-14B-720P-diffusers"
DTYPE        = torch.float16              # half-precision
MAX_AREA     = 1280 * 720                 # ≤720p
DEFAULT_FRAMES = 81                       # ≈5s @16fps
# --------------------------------------------------------------------

def load_pipeline():
    # 1) image encoder in full precision
    image_encoder = CLIPVisionModel.from_pretrained(
        MODEL_ID, subfolder="image_encoder", torch_dtype=torch.float32
    )
    # 2) VAE in reduced precision
    vae = AutoencoderKLWan.from_pretrained(
        MODEL_ID, subfolder="vae", torch_dtype=DTYPE
    )
    # 3) CLIPImageProcessor so we get the right class
    image_processor = CLIPImageProcessor.from_pretrained(
        MODEL_ID, subfolder="", torch_dtype=DTYPE
    )
    # 4) load everything with a balanced device map
    pipe = WanImageToVideoPipeline.from_pretrained(
        MODEL_ID,
        vae=vae,
        image_encoder=image_encoder,
        image_processor=image_processor,
        torch_dtype=DTYPE,
        device_map="balanced",           # splits weights CPU/GPU
    )
    return pipe

# load once at import
PIPE = load_pipeline()


# --------------------------------------------------------------------
# UTILS
def aspect_resize(img: Image.Image, max_area=MAX_AREA):
    """Resize while respecting multiples of the model’s patch size."""
    ar   = img.height / img.width
    mod  = PIPE.vae_scale_factor_spatial * PIPE.transformer.config.patch_size[1]
    h    = round(np.sqrt(max_area * ar)) // mod * mod
    w    = round(np.sqrt(max_area / ar)) // mod * mod
    return img.resize((w, h), Image.LANCZOS), h, w

def center_crop_resize(img: Image.Image, h, w):
    """Crop-and-resize to exactly (h, w)."""
    ratio = max(w / img.width, h / img.height)
    img   = img.resize(
        (round(img.width * ratio), round(img.height * ratio)),
        Image.LANCZOS
    )
    return TF.center_crop(img, [h, w])


# --------------------------------------------------------------------
# GENERATE (with simple progress streaming)
def generate(
    first_frame: Image.Image,
    last_frame: Image.Image,
    prompt: str,
    negative_prompt: str,
    steps: int,
    guidance: float,
    num_frames: int,
    seed: int,
    fps: int,
    progress=gr.Progress(),   # gradio’s built-in progress callback
):
    # pick or set seed
    if seed == -1:
        seed = torch.seed()
    gen = torch.Generator(device=PIPE.device).manual_seed(seed)

    # 0→10%: resize
    progress(0.0, desc="Resizing first frame…")
    first_frame, h, w = aspect_resize(first_frame)
    if last_frame.size != first_frame.size:
        progress(0.1, desc="Resizing last frame…")
        last_frame = center_crop_resize(last_frame, h, w)

    # 10→20%: ready to run
    progress(0.2, desc="Starting video inference…")
    result = PIPE(
        image=first_frame,
        last_image=last_frame,
        prompt=prompt,
        negative_prompt=negative_prompt or None,
        height=h,
        width=w,
        num_frames=num_frames,
        num_inference_steps=steps,
        guidance_scale=guidance,
        generator=gen,
    )

    # 80→100%: export
    progress(0.8, desc="Assembling video file…")
    video_path = export_to_video(result.frames[0], fps=fps)
    progress(1.0, desc="Done!")

    # return path so gr.File offers immediate download, plus seed used
    return video_path, seed


# --------------------------------------------------------------------
# UI
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown("## Wan2.1 FLF2V – First & Last Frame → Video")

    with gr.Row():
        first_img = gr.Image(label="First frame", type="pil")
        last_img  = gr.Image(label="Last frame",  type="pil")

    prompt          = gr.Textbox(label="Prompt", placeholder="A blue bird takes off…")
    negative        = gr.Textbox(label="Negative prompt (optional)", placeholder="ugly, blurry")

    with gr.Accordion("Advanced parameters", open=False):
        steps      = gr.Slider(10, 50, value=30, step=1, label="Sampling steps")
        guidance   = gr.Slider(0.0, 10.0, value=5.5, step=0.1, label="Guidance scale")
        num_frames = gr.Slider(16, 129, value=DEFAULT_FRAMES, step=1, label="Frames")
        fps        = gr.Slider(4, 30, value=16, step=1, label="FPS")
        seed       = gr.Number(value=-1, precision=0, label="Seed (-1=random)")

    run_btn     = gr.Button("Generate")
    download    = gr.File(label="Download video", interactive=False)
    used_seed   = gr.Number(label="Seed used", interactive=False)

    run_btn.click(
        fn=generate,
        inputs=[first_img, last_img, prompt, negative,
                steps, guidance, num_frames, seed, fps],
        outputs=[download, used_seed],
    )

# queue tasks so users see the little task-queue progress bar
demo.queue().launch(server_name="0.0.0.0", server_port=7860)