Spaces:
Runtime error
Runtime error
File size: 11,343 Bytes
836b387 33f757a 836b387 33f757a 836b387 33f757a 836b387 33f757a 836b387 33f757a 836b387 33f757a 836b387 33f757a 836b387 33f757a 836b387 33f757a 836b387 33f757a 836b387 d7d8851 9dbf912 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 |
#!/usr/bin/env python
from __future__ import annotations
import os
import gradio as gr
from webui.merge_config_gradio import merge_config_then_run
import huggingface_hub
import shutil
import os
import torch
HF_TOKEN = os.getenv('HF_TOKEN')
pipe = merge_config_then_run()
ARTICLE = r"""
If VideoGrain is helpful, please help to ⭐ the <a href='https://github.com/knightyxp/VideoGrain' target='_blank'>Github Repo</a>. Thanks!
[](https://github.com/knightyxp/VideoGrain)
---
📝 **Citation**
If our work is useful for your research, please consider citing:
```bibtex
@article{yang2025videograin,
title={VideoGrain: Modulating Space-Time Attention for Multi-grained Video Editing},
author={Yang, Xiangpeng and Zhu, Linchao and Fan, Hehe and Yang, Yi},
journal={ICLR},
year={2025}
}
```
📋 **License**
This project is licensed under <a rel="license" href="https://github.com/knightyxp/VideoGrain?tab=License-1-ov-file#readme">ReLER-Lab License 1.0</a>.
Redistribution and use for non-commercial purposes should follow this license.
📧 **Contact**
If you have any questions, please feel free to reach me out at <b>knightyxp@gmail.com</b>.
"""
def update_layout_visibility(selected_num):
num = int(selected_num)
return [gr.update(visible=(i < num)) for i in range(len(layout_files))]
with gr.Blocks(css='style.css') as demo:
# gr.Markdown(TITLE)
gr.HTML(
"""
<div style="text-align: center; max-width: 1200px; margin: 20px auto;">
<h1 style="font-weight: 900; font-size: 2rem; margin: 0rem">
VideoGrain: Modulating Space-Time Attention for Multi-Grained Video Editing
</h1>
<h2 style="font-weight: 450; font-size: 1rem; margin: 0rem">
<a href="https://github.com/knightyxp">Xiangpeng Yang</a>
</h2>
<h2 style="font-weight: 450; font-size: 1rem; margin: 0rem">
<span class="link-block">
[<a href="https://arxiv.org/abs/2502.17258" target="_blank"
class="external-link ">
<span class="icon">
<i class="ai ai-arxiv"></i>
</span>
<span>arXiv</span>
</a>]
</span>
<!-- Github link -->
<span class="link-block">
[<a href="https://github.com/knightyxp/VideoGrain" target="_blank"
class="external-link ">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code</span>
</a>]
</span>
<!-- Github link -->
<span class="link-block">
[<a href="https://knightyxp.github.io/VideoGrain_project_page" target="_blank"
class="external-link ">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Homepage</span>
</a>]
</span>
<!-- Github link -->
<span class="link-block">
[<a href="https://www.youtube.com/watch?v=XEM4Pex7F9E" target="_blank"
class="external-link ">
<span class="icon">
<i class="fab fa-youtube"></i>
</span>
<span>Youtube Video</span>
</a>]
</span>
</h2>
<h2 style="font-weight: 450; font-size: 1rem; margin-top: 0.5rem; margin-bottom: 0.5rem">
📕 TL;DR: VideoGrain is a zero-shot method for class-level, instance-level, and part-level video editing
</h2>
<h2 style="font-weight: 450; font-size: 1rem;">
Note that this page is a limited demo of VideoGrain. To run with more configurations, please check out our <a href="https://github.com/knightyxp/VideoGrain">github page.
</h2>
</div>
""")
gr.HTML("""
<p>We provide an <a href="https://github.com/knightyxp/VideoGrain?tab=readme-ov-file#editing-guidance-for-your-video"> Editing Guidance </a> to help users to choose hyperparameters when editing in-the-wild video.
<p>To remove the limitations or avoid queue on your own hardware, you may <a href="https://huggingface.co/spaces/XiangpengYang/VideoGrain?duplicate=true" style="display: inline-block; vertical-align: middle;"><img style="margin-top: 0em; margin-bottom: 0em; display: inline-block;" src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a></p>
""")
with gr.Row():
with gr.Column():
with gr.Accordion('Input Video', open=True):
# user_input_video = gr.File(label='Input Source Video')
user_input_video = gr.Video(label='Input Source Video', source='upload', type='numpy', format="mp4", visible=True).style(height="auto")
# Radio to choose how many layout videos to show
num_layouts = gr.Radio(
choices=["2", "3", "4", "5"],
label="Select Number of Editing Areas",
value="2", # default
info="Please select the number of editing areas"
)
# 使用循环生成所有的布局视频组件，并存到列表 layout_files 中
layout_files = []
with gr.Row():
for i in range(5):
video = gr.Video(
label=f"Layout Video {i+1}",
type="numpy",
format="mp4",
visible=(i < 2) # 默认显示前两个
)
layout_files.append(video)
# 当 num_layouts 改变时，通过回调函数更新 layout_files 列表中各视频组件的 visible 属性
num_layouts.change(
fn=update_layout_visibility,
inputs=num_layouts,
outputs=layout_files
)
prompt = gr.Textbox(label='Prompt',
info='Change the prompt, and extract each local prompt in the editing prompts.\
(the local prompt order should be same as layout masks order.)',
)
model_id = gr.Dropdown(
label='Model ID',
choices=[
'stable-diffusion-v1-5/stable-diffusion-v1-5',
# add shape editing ckpt here
],
value='stable-diffusion-v1-5/stable-diffusion-v1-5')
with gr.Column():
result = gr.Video(label='Result')
# result.style(height=512, width=512)
with gr.Accordion('Temporal Crop offset and Sampling Stride', open=False):
n_sample_frame = gr.Slider(label='Number of Frames',
minimum=0,
maximum=32,
step=1,
value=16)
sampling_rate = gr.Slider(label='sampling_rate',
minimum=0,
maximum=20,
step=1,
value=1)
start_sample_frame = gr.Number(label='Start frame in the video',
value=0,
precision=0)
with gr.Row():
control_list = ['dwpose', 'depth_zoe', 'depth_midas']
control_type = gr.Dropdown(
choices=control_list,
label='Control type',
value='dwpose'
)
# Checkbox group for "dwpose" options; default: hand selected, face not selected.
dwpose_options = gr.CheckboxGroup(
choices=["hand", "face"],
label="DW Pose Options",
value=["hand"],
visible=True # Initially visible since default control_type is "dwpose"
)
# Update the visibility of the dwpose_options based on the selected control type
control_type.change(
fn=lambda x: gr.update(visible=(x == "dwpose")),
inputs=control_type,
outputs=dwpose_options
)
controlnet_conditioning_scale = gr.Slider(label='ControlNet conditioning scale',
minimum=0.0,
maximum=1.0,
value=1.0,
step=0.1)
with gr.Accordion('Editing config for VideoGrian', open=True):
use_pnp = gr.Checkbox(
label="Use PnP",
value=False,
info="Check to enable PnP functionality."
)
pnp_inject_steps = gr.Slider(label='pnp inject steps',
info='PnP inject steps for temporal consistency',
minimum=0,
maximum=10,
step=1,
value=0)
flatten_res = gr.CheckboxGroup(
choices=["1", "2", "4", "8"],
label="Flatten Resolution",
value=["1"],
info="Select one or more flatten resolution factors. Mapping: 1 -> 64, 2 -> 32 (64/2), 4 -> 16 (64/4), 8 -> 8 (64/8)."
)
run_button = gr.Button('Generate')
with gr.Row():
from example import style_example
examples = style_example
# gr.Examples(examples=examples,
# inputs=[
# model_id,
# user_input_video,
# layout_files,
# prompt,
# model_id,
# control_type,
# dwpose_options,
# controlnet_conditioning_scale,
# use_pnp,
# pnp_inject_steps,
# flatten_res,
# ],
# outputs=result,
# fn=pipe.run,
# cache_examples=True,
# # cache_examples=os.getenv('SYSTEM') == 'spaces'
# )
gr.Markdown(ARTICLE)
inputs = [user_input_video, num_layouts,
*layout_files,
prompt,
model_id,
n_sample_frame,
start_sample_frame,
sampling_rate,
control_type,
dwpose_options,
controlnet_conditioning_scale,
use_pnp,
pnp_inject_steps,
flatten_res,
]
prompt.submit(fn=pipe.run, inputs=inputs, outputs=result)
run_button.click(fn=pipe.run, inputs=inputs, outputs=result)
demo.queue().launch(share=True) |