app.py

import torch
import gradio as gr
import spaces
import random
import os
from diffusers.utils import export_to_video
from diffusers import AutoencoderKLWan, WanPipeline
from diffusers.schedulers.scheduling_unipc_multistep import UniPCMultistepScheduler
from diffusers.schedulers.scheduling_flow_match_euler_discrete import FlowMatchEulerDiscreteScheduler
from huggingface_hub import hf_hub_download
from lycoris import create_lycoris_from_weights

# Define model options
MODEL_OPTIONS = {
    "Wan2.1-T2V-1.3B": "Wan-AI/Wan2.1-T2V-1.3B-Diffusers",
    "Wan2.1-T2V-14B": "Wan-AI/Wan2.1-T2V-14B-Diffusers"
}

# Define scheduler options
SCHEDULER_OPTIONS = {
    "UniPCMultistepScheduler": UniPCMultistepScheduler,
    "FlowMatchEulerDiscreteScheduler": FlowMatchEulerDiscreteScheduler
}

def download_adapter(repo_id, weight_name=None):
    """
    Download the adapter file from the Hugging Face Hub.
    If weight_name is not provided, attempts to use pytorch_lora_weights.safetensors
    """
    adapter_filename = weight_name if weight_name else "pytorch_lora_weights.safetensors"
    cache_dir = os.environ.get('HF_PATH', os.path.expanduser('~/.cache/huggingface/hub/models'))
    cleaned_adapter_path = repo_id.replace("/", "_").replace("\\", "_").replace(":", "_")
    path_to_adapter = os.path.join(cache_dir, cleaned_adapter_path)
    os.makedirs(path_to_adapter, exist_ok=True)
    
    try:
        path_to_adapter_file = hf_hub_download(
            repo_id=repo_id, 
            filename=adapter_filename, 
            local_dir=path_to_adapter
        )
        return path_to_adapter_file
    except Exception as e:
        # If specific file not found, try to get a list of available safetensors files
        if weight_name is None:
            raise ValueError(f"Could not download default adapter file: {str(e)}\nPlease specify the exact weight file name.")
        else:
            raise ValueError(f"Could not download adapter file {weight_name}: {str(e)}")

@spaces.GPU(duration=300)
def generate_video(
    model_choice,
    prompt,
    negative_prompt,
    lycoris_id,
    lycoris_weight_name,
    lycoris_scale,
    scheduler_type,
    flow_shift,
    height,
    width,
    num_frames,
    guidance_scale,
    num_inference_steps,
    output_fps,
    seed
):
    # Get model ID from selection
    model_id = MODEL_OPTIONS[model_choice]
    
    # Set seed for reproducibility
    if seed == -1 or seed is None or seed == "":
        seed = random.randint(0, 2147483647)
    else:
        seed = int(seed)
    
    # Set the seed
    torch.manual_seed(seed)
    
    # Load model
    vae = AutoencoderKLWan.from_pretrained(model_id, subfolder="vae", torch_dtype=torch.float32)
    pipe = WanPipeline.from_pretrained(model_id, vae=vae, torch_dtype=torch.float16)
    
    # Set scheduler
    if scheduler_type == "UniPCMultistepScheduler":
        pipe.scheduler = UniPCMultistepScheduler.from_config(
            pipe.scheduler.config,
            flow_shift=flow_shift
        )
    else:
        pipe.scheduler = FlowMatchEulerDiscreteScheduler(shift=flow_shift)
    
    # Move to GPU
    pipe.to("cuda")
    
    # Load LyCORIS weights if provided
    if lycoris_id and lycoris_id.strip():
        try:
            # Download the adapter file
            adapter_file_path = download_adapter(repo_id=lycoris_id, weight_name=lycoris_weight_name if lycoris_weight_name and lycoris_weight_name.strip() else None)
            
            # Apply LyCORIS adapter
            wrapper, *_ = create_lycoris_from_weights(lycoris_scale, adapter_file_path, pipe.transformer)
            wrapper.merge_to()
            
        except ValueError as e:
            # Return informative error if there are issues loading the adapter
            if "more than one weights file" in str(e) or "Could not download default adapter file" in str(e):
                return f"Error: The repository '{lycoris_id}' may contain multiple weight files. Please specify a weight name using the 'LyCORIS Weight Name' field.", seed
            else:
                return f"Error loading LyCORIS weights: {str(e)}", seed
    
    # Enable CPU offload for low VRAM
    pipe.enable_model_cpu_offload()
    
    # Generate video
    output = pipe(
        prompt=prompt,
        negative_prompt=negative_prompt,
        height=height,
        width=width,
        num_frames=num_frames,
        guidance_scale=guidance_scale,
        num_inference_steps=num_inference_steps,
        generator=torch.Generator("cuda").manual_seed(seed)
    ).frames[0]
    
    # Export to video
    temp_file = "output.mp4"
    export_to_video(output, temp_file, fps=output_fps)
    
    return temp_file, seed

# Create the Gradio interface
with gr.Blocks() as demo:
    
    gr.Markdown("# Wan 2.1 T2V")
    
    with gr.Row():
        with gr.Column(scale=1):
            model_choice = gr.Dropdown(
                choices=list(MODEL_OPTIONS.keys()),
                value="Wan2.1-T2V-1.3B",
                label="Model"
            )
            
            prompt = gr.Textbox(
                label="Prompt",
                value="",
                lines=3
            )
            
            negative_prompt = gr.Textbox(
                label="Negative Prompt",
                value="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿",
                lines=3
            )
            
            with gr.Row():
                lycoris_id = gr.Textbox(
                    label="Adapter Repo (e.g., markury/wan-st)",
                    value="markury/wan-st"
                )
            
            with gr.Row():
                lycoris_weight_name = gr.Textbox(
                    label="Adapter Path in Repo",
                    value="pytorch_lora_weights.safetensors",
                    info="Specify for repos with multiple .safetensors files, e.g.: adapter_model.safetensors, pytorch_lora_weights.safetensors, etc."
                )
                lycoris_scale = gr.Slider(
                    label="Adapter Scale",
                    minimum=0.0,
                    maximum=2.0,
                    value=1.00,
                    step=0.05
                )
            
            with gr.Row():
                scheduler_type = gr.Dropdown(
                    choices=list(SCHEDULER_OPTIONS.keys()),
                    value="UniPCMultistepScheduler",
                    label="Scheduler"
                )
                flow_shift = gr.Slider(
                    label="Flow Shift",
                    minimum=1.0,
                    maximum=12.0,
                    value=3.0,
                    step=0.5,
                    info="2.0-5.0 for smaller videos, 7.0-12.0 for larger videos"
                )
            
            with gr.Row():
                height = gr.Slider(
                    label="Height",
                    minimum=256,
                    maximum=1024,
                    value=832,
                    step=32
                )
                width = gr.Slider(
                    label="Width",
                    minimum=256,
                    maximum=1792,
                    value=480,
                    step=30
                )
            
            with gr.Row():
                num_frames = gr.Slider(
                    label="Number of Frames (4k+1 is recommended, e.g. 33)",
                    minimum=17,
                    maximum=129,
                    value=33,
                    step=4
                )
                output_fps = gr.Slider(
                    label="Output FPS",
                    minimum=8,
                    maximum=30,
                    value=16,
                    step=1
                )
            
            with gr.Row():
                guidance_scale = gr.Slider(
                    label="Guidance Scale (CFG)",
                    minimum=1.0,
                    maximum=15.0,
                    value=4.0,
                    step=0.5
                )
                num_inference_steps = gr.Slider(
                    label="Inference Steps",
                    minimum=10,
                    maximum=100,
                    value=20,
                    step=1
                )
            
            seed = gr.Number(
                label="Seed (-1 for random)",
                value=-1,
                precision=0,
                info="Set a specific seed for deterministic results"
            )
            
            generate_btn = gr.Button("Generate Video")
        
        with gr.Column(scale=1):
            output_video = gr.Video(label="Generated Video")
            used_seed = gr.Number(label="Seed", precision=0)
            
    generate_btn.click(
        fn=generate_video,
        inputs=[
            model_choice,
            prompt,
            negative_prompt,
            lycoris_id,
            lycoris_weight_name,
            lycoris_scale,
            scheduler_type,
            flow_shift,
            height,
            width,
            num_frames,
            guidance_scale,
            num_inference_steps,
            output_fps,
            seed
        ],
        outputs=[output_video, used_seed]
    )
    
    gr.Markdown("""
    ## Tips for best results:
    - For smaller resolution videos, try lower values of flow shift (2.0-5.0)
    - For larger resolution videos, try higher values of flow shift (7.0-12.0)
    - Number of frames should be of the form 4k+1 (e.g., 33, 81)
    - Stick to lower frame counts. Even at 480p, an 81 frame sequence at 30 steps will nearly time out the request in this ZeroGPU space.
    
    """)

demo.launch()