Spaces:
Running
on
Zero
Running
on
Zero
File size: 10,128 Bytes
fc39e7c dabe886 89892f2 637f73a dabe886 fc39e7c dabe886 fc39e7c dabe886 40667cc fc39e7c dabe886 40667cc dabe886 fc39e7c 637f73a 362f999 fc39e7c |
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 |
# coding: utf-8
"""
The entrance of the gradio
"""
import tyro
import gradio as gr
import os.path as osp
from src.utils.helper import load_description
from src.gradio_pipeline import GradioPipeline
from src.config.crop_config import CropConfig
from src.config.argument_config import ArgumentConfig
from src.config.inference_config import InferenceConfig
import spaces
import cv2
#추가
from elevenlabs_utils import ElevenLabsPipeline
from setup_environment import initialize_environment
from src.utils.video import extract_audio
from flux_dev import create_flux_tab
# import gdown
# folder_url = f"https://drive.google.com/drive/folders/1UtKgzKjFAOmZkhNK-OYT0caJ_w2XAnib"
# gdown.download_folder(url=folder_url, output="pretrained_weights", quiet=False)
# import sys
# sys.path.append('/home/user/.local/lib/python3.10/site-packages')
# sys.path.append('/home/user/.local/lib/python3.10/site-packages/stf_alternative/src/stf_alternative')
# sys.path.append('/home/user/.local/lib/python3.10/site-packages/stf_tools/src/stf_tools')
# sys.path.append('/home/user/app/')
# sys.path.append('/home/user/app/stf/')
# sys.path.append('/home/user/app/stf/stf_alternative/')
# sys.path.append('/home/user/app/stf/stf_alternative/src/stf_alternative')
# sys.path.append('/home/user/app/stf/stf_tools')
# sys.path.append('/home/user/app/stf/stf_tools/src/stf_tools')
# # CUDA 경로를 환경 변수로 설정
# os.environ['PATH'] = '/usr/local/cuda/bin:' + os.environ.get('PATH', '')
# os.environ['LD_LIBRARY_PATH'] = '/usr/local/cuda/lib64:' + os.environ.get('LD_LIBRARY_PATH', '')
# # 확인용 출력
# print("PATH:", os.environ['PATH'])
# print("LD_LIBRARY_PATH:", os.environ['LD_LIBRARY_PATH'])
# from stf_utils import STFPipeline
def partial_fields(target_class, kwargs):
return target_class(**{k: v for k, v in kwargs.items() if hasattr(target_class, k)})
# set tyro theme
tyro.extras.set_accent_color("bright_cyan")
args = tyro.cli(ArgumentConfig)
# specify configs for inference
inference_cfg = partial_fields(InferenceConfig, args.__dict__) # use attribute of args to initial InferenceConfig
crop_cfg = partial_fields(CropConfig, args.__dict__) # use attribute of args to initial CropConfig
gradio_pipeline = GradioPipeline(
inference_cfg=inference_cfg,
crop_cfg=crop_cfg,
args=args
)
# 추가 정의
elevenlabs_pipeline = ElevenLabsPipeline()
@spaces.GPU(duration=200)
def gpu_wrapped_elevenlabs_pipeline_generate_voice(text, voice):
return elevenlabs_pipeline.generate_voice(text, voice)
@spaces.GPU(duration=240)
def gpu_wrapped_execute_video(*args, **kwargs):
return gradio_pipeline.execute_video(*args, **kwargs)
@spaces.GPU(duration=240)
def gpu_wrapped_execute_image(*args, **kwargs):
return gradio_pipeline.execute_image(*args, **kwargs)
def is_square_video(video_path):
video = cv2.VideoCapture(video_path)
width = int(video.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(video.get(cv2.CAP_PROP_FRAME_HEIGHT))
video.release()
if width != height:
raise gr.Error("Error: the video does not have a square aspect ratio. We currently only support square videos")
return gr.update(visible=True)
# assets
title_md = "assets/gradio_title.md"
example_portrait_dir = "assets/examples/source"
example_video_dir = "assets/examples/driving"
data_examples = [
[osp.join(example_portrait_dir, "s9.jpg"), osp.join(example_video_dir, "d0.mp4"), True, True, True, True],
[osp.join(example_portrait_dir, "s6.jpg"), osp.join(example_video_dir, "d0.mp4"), True, True, True, True],
[osp.join(example_portrait_dir, "s10.jpg"), osp.join(example_video_dir, "d0.mp4"), True, True, True, True],
[osp.join(example_portrait_dir, "s5.jpg"), osp.join(example_video_dir, "d18.mp4"), True, True, True, True],
[osp.join(example_portrait_dir, "s7.jpg"), osp.join(example_video_dir, "d19.mp4"), True, True, True, True],
[osp.join(example_portrait_dir, "s22.jpg"), osp.join(example_video_dir, "d0.mp4"), True, True, True, True],
]
#################### interface logic ####################
# Define components first
eye_retargeting_slider = gr.Slider(minimum=0, maximum=0.8, step=0.01, label="target eyes-open ratio")
lip_retargeting_slider = gr.Slider(minimum=0, maximum=0.8, step=0.01, label="target lip-open ratio")
retargeting_input_image = gr.Image(type="filepath")
output_image = gr.Image(type="numpy")
output_image_paste_back = gr.Image(type="numpy")
output_video = gr.Video()
output_video_concat = gr.Video()
with gr.Blocks(theme=gr.themes.Soft()) as demo:
#gr.HTML(load_description(title_md))
with gr.Tabs():
with gr.Tab("Text to LipSync"):
gr.Markdown("# Text to LipSync")
with gr.Row():
with gr.Column():
script_txt = gr.Text()
with gr.Column():
audio_gen_button = gr.Button("Audio generation", variant="primary")
with gr.Row():
output_audio_path = gr.Audio(label="Generated audio", type="filepath")
gr.Markdown(load_description("assets/gradio_description_upload.md"))
with gr.Row():
with gr.Accordion(open=True, label="Source Portrait"):
image_input = gr.Image(type="filepath")
gr.Examples(
examples=[
[osp.join(example_portrait_dir, "s9.jpg")],
[osp.join(example_portrait_dir, "s6.jpg")],
[osp.join(example_portrait_dir, "s10.jpg")],
[osp.join(example_portrait_dir, "s5.jpg")],
[osp.join(example_portrait_dir, "s7.jpg")],
[osp.join(example_portrait_dir, "s12.jpg")],
[osp.join(example_portrait_dir, "s22.jpg")],
],
inputs=[image_input],
cache_examples=False,
)
with gr.Accordion(open=True, label="Driving Video"):
video_input = gr.Video()
gr.Examples(
examples=[
[osp.join(example_video_dir, "d0.mp4")],
[osp.join(example_video_dir, "d18.mp4")],
[osp.join(example_video_dir, "d19.mp4")],
[osp.join(example_video_dir, "d14_trim.mp4")],
[osp.join(example_video_dir, "d6_trim.mp4")],
],
inputs=[video_input],
cache_examples=False,
)
with gr.Row():
with gr.Accordion(open=False, label="Animation Instructions and Options"):
gr.Markdown(load_description("assets/gradio_description_animation.md"))
with gr.Row():
flag_relative_input = gr.Checkbox(value=True, label="relative motion")
flag_do_crop_input = gr.Checkbox(value=True, label="do crop")
flag_remap_input = gr.Checkbox(value=True, label="paste-back")
gr.Markdown(load_description("assets/gradio_description_animate_clear.md"))
with gr.Row():
with gr.Column():
process_button_animation = gr.Button("🚀 Animate", variant="primary")
with gr.Column():
process_button_reset = gr.ClearButton([image_input, video_input, output_video, output_video_concat], value="🧹 Clear")
with gr.Row():
with gr.Column():
with gr.Accordion(open=True, label="The animated video in the original image space"):
output_video.render()
with gr.Column():
with gr.Accordion(open=True, label="The animated video"):
output_video_concat.render()
with gr.Row():
# Examples
gr.Markdown("## You could also choose the examples below by one click ⬇️")
with gr.Row():
gr.Examples(
examples=data_examples,
fn=gpu_wrapped_execute_video,
inputs=[
image_input,
video_input,
flag_relative_input,
flag_do_crop_input,
flag_remap_input
],
outputs=[output_image, output_image_paste_back],
examples_per_page=6,
cache_examples=False,
)
process_button_animation.click(
fn=gpu_wrapped_execute_video,
inputs=[
image_input,
video_input,
flag_relative_input,
flag_do_crop_input,
flag_remap_input
],
outputs=[output_video, output_video_concat],
show_progress=True
)
audio_gen_button.click(
fn=gpu_wrapped_elevenlabs_pipeline_generate_voice,
inputs=[
script_txt
],
outputs=[output_audio_path],
show_progress=True
)
# image_input.change(
# fn=gradio_pipeline.prepare_retargeting,
# inputs=image_input,
# outputs=[eye_retargeting_slider, lip_retargeting_slider, retargeting_input_image]
# )
video_input.upload(
fn=is_square_video,
inputs=video_input,
outputs=video_input
)
# 세 번째 탭: Flux 개발용 탭
with gr.Tab("FLUX Dev"):
flux_demo = create_flux_tab(image_input) # Flux 개발용 탭 생성
demo.launch(
server_port=args.server_port,
share=args.share,
server_name=args.server_name
)
|