app.py

import gradio as gr

import os
import sys
import argparse
import random
from omegaconf import OmegaConf
import torch
import torchvision
from pytorch_lightning import seed_everything
from huggingface_hub import hf_hub_download

sys.path.insert(0, "scripts/evaluation")
from funcs import (
    batch_ddim_sampling_freenoise,
    load_model_checkpoint,
)
from utils.utils import instantiate_from_config

def infer(prompt, output_size, seed, num_frames, ddim_steps, unconditional_guidance_scale, save_fps):
    window_size = 16
    window_stride = 4
    
    if output_size == "320x512":
        width = 512 
        height = 320
        ckpt_dir_512 = "checkpoints/base_512_v2"
        ckpt_path_512 = "checkpoints/base_512_v2/model.ckpt"
        config_512 = "configs/inference_t2v_tconv512_v2.0_freenoise.yaml"
        config_512 = OmegaConf.load(config_512)
        model_config_512 = config_512.pop("model", OmegaConf.create())
        model_512 = instantiate_from_config(model_config_512)
        model_512 = model_512.cuda()
        if not os.path.exists(ckpt_path_512):
            os.makedirs(ckpt_dir_512, exist_ok=True)
            hf_hub_download(repo_id="VideoCrafter/VideoCrafter2", filename="model.ckpt", local_dir=ckpt_dir_512)
        try:
            model_512 = load_model_checkpoint(model_512, ckpt_path_512)
        except:
            hf_hub_download(repo_id="VideoCrafter/VideoCrafter2", filename="model.ckpt", local_dir=ckpt_dir_512, force_download=True)
            model_512 = load_model_checkpoint(model_512, ckpt_path_512)
        model_512.eval()
        model = model_512
        fps = 12
    if output_size == "576x1024":
        width = 1024
        height = 576
        ckpt_dir_1024 = "checkpoints/base_1024_v1"
        ckpt_path_1024 = "checkpoints/base_1024_v1/model.ckpt"
        config_1024 = "configs/inference_t2v_1024_v1.0_freenoise.yaml"
        config_1024 = OmegaConf.load(config_1024)
        model_config_1024 = config_1024.pop("model", OmegaConf.create())
        model_1024 = instantiate_from_config(model_config_1024)
        model_1024 = model_1024.cuda()
        if not os.path.exists(ckpt_path_1024):
            os.makedirs(ckpt_dir_1024, exist_ok=True)
            hf_hub_download(repo_id="VideoCrafter/Text2Video-1024", filename="model.ckpt", local_dir=ckpt_dir_1024)
        try:
            model_1024 = load_model_checkpoint(model_1024, ckpt_path_1024)
        except:
            hf_hub_download(repo_id="VideoCrafter/Text2Video-1024", filename="model.ckpt", local_dir=ckpt_dir_1024, force_download=True)
            model_1024 = load_model_checkpoint(model_1024, ckpt_path_1024)
        model_1024.eval()
        model = model_1024
        fps = 28
        num_frames = min(num_frames, 36)
    elif output_size == "256x256":
        width = 256 
        height = 256
        ckpt_dir_256 = "checkpoints/base_256_v1"
        ckpt_path_256 = "checkpoints/base_256_v1/model.ckpt"
        config_256 = "configs/inference_t2v_tconv256_v1.0_freenoise.yaml"
        config_256 = OmegaConf.load(config_256)
        model_config_256 = config_256.pop("model", OmegaConf.create())
        model_256 = instantiate_from_config(model_config_256)
        model_256 = model_256.cuda()
        if not os.path.exists(ckpt_path_256):
            os.makedirs(ckpt_dir_256, exist_ok=True)
            hf_hub_download(repo_id="VideoCrafter/Text2Video-256", filename="model.ckpt", local_dir=ckpt_dir_256)
        try:
            model_256 = load_model_checkpoint(model_256, ckpt_path_256)
        except:
            hf_hub_download(repo_id="VideoCrafter/Text2Video-256", filename="model.ckpt", local_dir=ckpt_dir_256, force_download=True)
            model_256 = load_model_checkpoint(model_256, ckpt_path_256)
        model_256.eval()
        model = model_256
        fps = 8

    if seed is None:
        seed = int.from_bytes(os.urandom(2), "big")
    print(f"Using seed: {seed}")
    seed_everything(seed)

    args = argparse.Namespace(
        mode="base",
        savefps=save_fps,
        n_samples=1,
        ddim_steps=ddim_steps,
        ddim_eta=0.0,
        bs=1,
        height=height,
        width=width,
        frames=num_frames,
        fps=fps,
        unconditional_guidance_scale=unconditional_guidance_scale,
        unconditional_guidance_scale_temporal=None,
        cond_input=None,
        window_size=window_size,
        window_stride=window_stride,
    )

    ## latent noise shape
    h, w = args.height // 8, args.width // 8
    frames = model.temporal_length if args.frames < 0 else args.frames
    channels = model.channels

    x_T_total = torch.randn(
        [args.n_samples, 1, channels, frames, h, w], device=model.device
    ).repeat(1, args.bs, 1, 1, 1, 1)
    for frame_index in range(args.window_size, args.frames, args.window_stride):
        list_index = list(
            range(
                frame_index - args.window_size,
                frame_index + args.window_stride - args.window_size,
            )
        )
        random.shuffle(list_index)
        x_T_total[
            :, :, :, frame_index : frame_index + args.window_stride
        ] = x_T_total[:, :, :, list_index]

    batch_size = 1
    noise_shape = [batch_size, channels, frames, h, w]
    fps = torch.tensor([args.fps] * batch_size).to(model.device).long()
    prompts = [prompt]
    text_emb = model.get_learned_conditioning(prompts)

    cond = {"c_crossattn": [text_emb], "fps": fps}

    ## inference
    batch_samples = batch_ddim_sampling_freenoise(
        model,
        cond,
        noise_shape,
        args.n_samples,
        args.ddim_steps,
        args.ddim_eta,
        args.unconditional_guidance_scale,
        args=args,
        x_T_total=x_T_total,
    )

    video_path = "output.mp4"
    vid_tensor = batch_samples[0]
    video = vid_tensor.detach().cpu()
    video = torch.clamp(video.float(), -1.0, 1.0)
    video = video.permute(2, 0, 1, 3, 4)  # t,n,c,h,w

    frame_grids = [
        torchvision.utils.make_grid(framesheet, nrow=int(args.n_samples))
        for framesheet in video
    ]  # [3, 1*h, n*w]
    grid = torch.stack(frame_grids, dim=0)  # stack in temporal dim [t, 3, n*h, w]
    grid = (grid + 1.0) / 2.0
    grid = (grid * 255).to(torch.uint8).permute(0, 2, 3, 1)

    torchvision.io.write_video(
        video_path,
        grid,
        fps=args.savefps,
        video_codec="h264",
        options={"crf": "10"},
    )
    
    print(video_path)
    return video_path

examples = [
    ["A chihuahua in astronaut suit floating in space, cinematic lighting, glow effect",],
    ["A corgi is swimming quickly",],
    ["A bigfoot walking in the snowstorm",],
    ["Campfire at night in a snowy forest with starry sky in the background",],
    ["A panda is surfing in the universe",],
]

css = """
#col-container {max-width: 640px; margin-left: auto; margin-right: auto;}
a {text-decoration-line: underline; font-weight: 600;}
.animate-spin {
  animation: spin 1s linear infinite;
}
@keyframes spin {
  from {
      transform: rotate(0deg);
  }
  to {
      transform: rotate(360deg);
  }
}
#share-btn-container {
  display: flex; 
  padding-left: 0.5rem !important; 
  padding-right: 0.5rem !important; 
  background-color: #000000; 
  justify-content: center; 
  align-items: center; 
  border-radius: 9999px !important; 
  max-width: 15rem;
  height: 36px;
}
div#share-btn-container > div {
    flex-direction: row;
    background: black;
    align-items: center;
}
#share-btn-container:hover {
  background-color: #060606;
}
#share-btn {
  all: initial; 
  color: #ffffff;
  font-weight: 600; 
  cursor:pointer; 
  font-family: 'IBM Plex Sans', sans-serif; 
  margin-left: 0.5rem !important; 
  padding-top: 0.5rem !important; 
  padding-bottom: 0.5rem !important;
  right:0;
}
#share-btn * {
  all: unset;
}
#share-btn-container div:nth-child(-n+2){
  width: auto !important;
  min-height: 0px !important;
}
#share-btn-container .wrap {
  display: none !important;
}
#share-btn-container.hidden {
  display: none!important;
}
img[src*='#center'] { 
    display: inline-block;
    margin: unset;
}
.footer {
        margin-bottom: 45px;
        margin-top: 10px;
        text-align: center;
        border-bottom: 1px solid #e5e5e5;
    }
    .footer>p {
        font-size: .8rem;
        display: inline-block;
        padding: 0 10px;
        transform: translateY(10px);
        background: white;
    }
    .dark .footer {
        border-color: #303030;
    }
    .dark .footer>p {
        background: #0b0f19;
    }
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown(
            """
            <h1 style="text-align: center;">FreeNoise (Longer Text-to-Video)</h1>
            <p style="text-align: center;">
            FreeNoise: Tuning-Free Longer Video Diffusion via Noise Rescheduling (ICLR 2024)
            </p>
            <p style="text-align: center;">
            <a href="https://arxiv.org/abs/2310.15169" target="_blank"><b>[arXiv]</b></a> &nbsp;&nbsp;&nbsp;&nbsp;
            <a href="http://haonanqiu.com/projects/FreeNoise.html" target="_blank"><b>[Project Page]</b></a> &nbsp;&nbsp;&nbsp;&nbsp;
            <a href="https://github.com/AILab-CVC/FreeNoise" target="_blank"><b>[Code]</b></a>
            </p>         
            """
        )

        prompt_in = gr.Textbox(label="Prompt", placeholder="A chihuahua in astronaut suit floating in space, cinematic lighting, glow effect")

        with gr.Row():
            with gr.Accordion('FreeNoise Parameters (feel free to adjust these parameters based on your prompt): ', open=False):
                with gr.Row():
                    output_size = gr.Dropdown(["320x512", "576x1024", "256x256"], value="320x512", label="Output Size", info="250s for 512 model, 900s for 1024 model (32 frames). Recovering from sleeping will take more time to download ckpt")
                with gr.Row():
                    num_frames = gr.Slider(label='Frames (a multiple of 4), max 36 for 1024 model',
                             minimum=16,
                             maximum=64,
                             step=4,
                             value=32)
                    ddim_steps = gr.Slider(label='DDIM Steps',
                             minimum=5,
                             maximum=200,
                             step=1,
                             value=50)
                with gr.Row():
                    unconditional_guidance_scale = gr.Slider(label='Unconditional Guidance Scale',
                             minimum=1.0,
                             maximum=20.0,
                             step=0.1,
                             value=12.0)
                    save_fps = gr.Slider(label='Save FPS',
                             minimum=1,
                             maximum=30,
                             step=1,
                             value=10)
                with gr.Row():
                    seed = gr.Slider(label='Random Seed',
                             minimum=0,
                             maximum=10000,
                             step=1,
                             value=123)

        submit_btn = gr.Button("Generate", variant='primary')
        video_result = gr.Video(label="Video Output")

        gr.Examples(examples=examples, inputs=[prompt_in, output_size, seed, num_frames, ddim_steps, unconditional_guidance_scale, save_fps])

    submit_btn.click(fn=infer,
            inputs=[prompt_in, output_size, seed, num_frames, ddim_steps, unconditional_guidance_scale, save_fps],
            outputs=[video_result],
            api_name="zrscp")

demo.queue(max_size=12).launch(show_api=True)