Spaces:
Running
on
Zero
Running
on
Zero
File size: 10,205 Bytes
967924e dbac20f c4dd2de 0cdb73b d2a875e 59b0bed c4dd2de dbac20f d2a875e 59b0bed dbac20f 59b0bed dbac20f d2a875e dbac20f d2a875e 59b0bed 0cdb73b 1354a37 d2a875e dbac20f 59b0bed dbac20f 59b0bed dbac20f 59b0bed dbac20f ba6e005 d2a875e 59b0bed ba6e005 59b0bed d2a875e ba6e005 59b0bed 35c2752 59b0bed dbac20f 77e7170 dbac20f d2a875e dbac20f 9ac63db dbac20f 627e0b8 dbac20f d2a875e dbac20f 9ac63db dbac20f ce31a26 d2a875e ce31a26 59b0bed d2a875e dbac20f ce31a26 dbac20f ce31a26 d2a875e dbac20f ce31a26 dbac20f ce31a26 d2a875e dbac20f ce31a26 dbac20f d2a875e ce31a26 d2a875e ce31a26 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 |
import spaces
import logging
from datetime import datetime
from pathlib import Path
import gradio as gr
import torch
import torchaudio
import os
import requests
from transformers import pipeline
import tempfile
import numpy as np
from einops import rearrange
import cv2
from scipy.io import wavfile
import librosa
import json
from typing import Optional, Tuple, List
import atexit
try:
import mmaudio
except ImportError:
os.system("pip install -e .")
import mmaudio
from mmaudio.eval_utils import (ModelConfig, all_model_cfg, generate, load_video, make_video,
setup_eval_logging)
from mmaudio.model.flow_matching import FlowMatching
from mmaudio.model.networks import MMAudio, get_my_mmaudio
from mmaudio.model.sequence_config import SequenceConfig
from mmaudio.model.utils.features_utils import FeaturesUtils
# ๋ก๊น
์ค์
logging.basicConfig(
level=logging.INFO,
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
)
log = logging.getLogger()
# CUDA ์ค์
if torch.cuda.is_available():
device = torch.device("cuda")
torch.backends.cuda.matmul.allow_tf32 = True
torch.backends.cudnn.allow_tf32 = True
torch.backends.cudnn.benchmark = True
else:
device = torch.device("cpu")
dtype = torch.bfloat16
# ๋ชจ๋ธ ์ค์
model: ModelConfig = all_model_cfg['large_44k_v2']
model.download_if_needed()
output_dir = Path('./output/gradio')
setup_eval_logging()
# ๋ฒ์ญ๊ธฐ ๋ฐ Pixabay API ์ค์
translator = pipeline("translation", model="Helsinki-NLP/opus-mt-ko-en", device="cpu")
PIXABAY_API_KEY = "33492762-a28a596ec4f286f84cd328b17"
def cleanup_temp_files():
temp_dir = tempfile.gettempdir()
for file in os.listdir(temp_dir):
if file.endswith(('.mp4', '.flac')):
try:
os.remove(os.path.join(temp_dir, file))
except:
pass
atexit.register(cleanup_temp_files)
def get_model() -> tuple[MMAudio, FeaturesUtils, SequenceConfig]:
with torch.cuda.device(device):
seq_cfg = model.seq_cfg
net: MMAudio = get_my_mmaudio(model.model_name).to(device, dtype).eval()
net.load_weights(torch.load(model.model_path, map_location=device, weights_only=True))
log.info(f'Loaded weights from {model.model_path}')
feature_utils = FeaturesUtils(
tod_vae_ckpt=model.vae_path,
synchformer_ckpt=model.synchformer_ckpt,
enable_conditions=True,
mode=model.mode,
bigvgan_vocoder_ckpt=model.bigvgan_16k_path,
need_vae_encoder=False
).to(device, dtype).eval()
return net, feature_utils, seq_cfg
net, feature_utils, seq_cfg = get_model()
# search_videos ํจ์ ์์
@torch.no_grad()
def search_videos(query):
try:
# CPU์์ ๋ฒ์ญ ์คํ
query = translate_prompt(query)
return search_pixabay_videos(query, PIXABAY_API_KEY)
except Exception as e:
logging.error(f"Video search error: {e}")
return []
# translate_prompt ํจ์๋ ์์
def translate_prompt(text):
try:
if text and any(ord(char) >= 0x3131 and ord(char) <= 0xD7A3 for char in text):
# CPU์์ ๋ฒ์ญ ์คํ
with torch.no_grad():
translation = translator(text)[0]['translation_text']
return translation
return text
except Exception as e:
logging.error(f"Translation error: {e}")
return text
# ๋๋ฐ์ด์ค ์ค์ ๋ถ๋ถ ์์
if torch.cuda.is_available():
device = torch.device("cuda")
torch.backends.cuda.matmul.allow_tf32 = True
torch.backends.cudnn.allow_tf32 = True
torch.backends.cudnn.benchmark = True
else:
device = torch.device("cpu")
# ๋ฒ์ญ๊ธฐ ์ค์ ์์
translator = pipeline("translation",
model="Helsinki-NLP/opus-mt-ko-en",
device="cpu") # ๋ช
์์ ์ผ๋ก CPU ์ง์
def search_pixabay_videos(query, api_key):
try:
base_url = "https://pixabay.com/api/videos/"
params = {
"key": api_key,
"q": query,
"per_page": 40
}
response = requests.get(base_url, params=params)
if response.status_code == 200:
data = response.json()
return [video['videos']['large']['url'] for video in data.get('hits', [])]
return []
except Exception as e:
logging.error(f"Pixabay API error: {e}")
return []
@spaces.GPU
@torch.inference_mode()
def video_to_audio(video: gr.Video, prompt: str, negative_prompt: str, seed: int, num_steps: int,
cfg_strength: float, duration: float):
prompt = translate_prompt(prompt)
negative_prompt = translate_prompt(negative_prompt)
rng = torch.Generator(device=device)
rng.manual_seed(seed)
fm = FlowMatching(min_sigma=0, inference_mode='euler', num_steps=num_steps)
clip_frames, sync_frames, duration = load_video(video, duration)
clip_frames = clip_frames.unsqueeze(0)
sync_frames = sync_frames.unsqueeze(0)
seq_cfg.duration = duration
net.update_seq_lengths(seq_cfg.latent_seq_len, seq_cfg.clip_seq_len, seq_cfg.sync_seq_len)
audios = generate(clip_frames,
sync_frames, [prompt],
negative_text=[negative_prompt],
feature_utils=feature_utils,
net=net,
fm=fm,
rng=rng,
cfg_strength=cfg_strength)
audio = audios.float().cpu()[0]
video_save_path = tempfile.NamedTemporaryFile(delete=False, suffix='.mp4').name
make_video(video,
video_save_path,
audio,
sampling_rate=seq_cfg.sampling_rate,
duration_sec=seq_cfg.duration)
return video_save_path
@spaces.GPU
@torch.inference_mode()
def text_to_audio(prompt: str, negative_prompt: str, seed: int, num_steps: int, cfg_strength: float,
duration: float):
prompt = translate_prompt(prompt)
negative_prompt = translate_prompt(negative_prompt)
rng = torch.Generator(device=device)
rng.manual_seed(seed)
fm = FlowMatching(min_sigma=0, inference_mode='euler', num_steps=num_steps)
clip_frames = sync_frames = None
seq_cfg.duration = duration
net.update_seq_lengths(seq_cfg.latent_seq_len, seq_cfg.clip_seq_len, seq_cfg.sync_seq_len)
audios = generate(clip_frames,
sync_frames, [prompt],
negative_text=[negative_prompt],
feature_utils=feature_utils,
net=net,
fm=fm,
rng=rng,
cfg_strength=cfg_strength)
audio = audios.float().cpu()[0]
audio_save_path = tempfile.NamedTemporaryFile(delete=False, suffix='.flac').name
torchaudio.save(audio_save_path, audio, seq_cfg.sampling_rate)
return audio_save_path
# CSS ์คํ์ผ ์์
custom_css = """
.gradio-container {
background: linear-gradient(45deg, #1a1a1a, #2a2a2a);
border-radius: 15px;
box-shadow: 0 8px 32px rgba(0,0,0,0.3);
color: #e0e0e0;
}
.input-container, .output-container {
background: rgba(40, 40, 40, 0.95);
backdrop-filter: blur(10px);
border-radius: 10px;
padding: 20px;
transform-style: preserve-3d;
transition: transform 0.3s ease;
border: 1px solid rgba(255, 255, 255, 0.1);
}
.input-container:hover {
transform: translateZ(20px);
box-shadow: 0 8px 32px rgba(0,0,0,0.5);
}
.gallery-item {
transition: transform 0.3s ease;
border-radius: 8px;
overflow: hidden;
background: #2a2a2a;
}
.gallery-item:hover {
transform: scale(1.05);
box-shadow: 0 4px 15px rgba(0,0,0,0.4);
}
.tabs {
background: rgba(30, 30, 30, 0.95);
border-radius: 10px;
padding: 10px;
border: 1px solid rgba(255, 255, 255, 0.05);
}
button {
background: linear-gradient(45deg, #2196F3, #1976D2);
border: none;
border-radius: 5px;
transition: all 0.3s ease;
color: white;
}
button:hover {
transform: translateY(-2px);
box-shadow: 0 4px 15px rgba(33,150,243,0.3);
}
/* ํ
์คํธ ์
๋ ฅ ํ๋ ์คํ์ผ */
textarea, input[type="text"], input[type="number"] {
background: rgba(30, 30, 30, 0.95) !important;
border: 1px solid rgba(255, 255, 255, 0.1) !important;
color: #e0e0e0 !important;
border-radius: 5px !important;
}
/* ๋ ์ด๋ธ ์คํ์ผ */
label {
color: #e0e0e0 !important;
}
/* ๊ฐค๋ฌ๋ฆฌ ๊ทธ๋ฆฌ๋ ์คํ์ผ */
.gallery {
background: rgba(30, 30, 30, 0.95);
padding: 15px;
border-radius: 10px;
border: 1px solid rgba(255, 255, 255, 0.05);
}
"""
# ์ธํฐํ์ด์ค ์ ์ ์์ (์๋ฌธ์ผ๋ก ๋ณ๊ฒฝ)
video_search_tab = gr.Interface(
fn=search_videos,
inputs=gr.Textbox(label="Search Query(ํ๊ธ์ง์)"),
outputs=gr.Gallery(label="Search Results", columns=4, rows=20),
css=custom_css,
api_name=False
)
video_to_audio_tab = gr.Interface(
fn=video_to_audio,
inputs=[
gr.Video(label="Input Video"),
gr.Textbox(label="Prompt(ํ๊ธ์ง์)"),
gr.Textbox(label="Negative Prompt", value="music"),
gr.Number(label="Seed", value=0),
gr.Number(label="Steps", value=25),
gr.Number(label="Guidance Scale", value=4.5),
gr.Number(label="Duration (sec)", value=8),
],
outputs=gr.Video(label="Generated Result"),
css=custom_css
)
text_to_audio_tab = gr.Interface(
fn=text_to_audio,
inputs=[
gr.Textbox(label="Prompt(ํ๊ธ์ง์)"),
gr.Textbox(label="Negative Prompt"),
gr.Number(label="Seed", value=0),
gr.Number(label="Steps", value=25),
gr.Number(label="Guidance Scale", value=4.5),
gr.Number(label="Duration (sec)", value=8),
],
outputs=gr.Audio(label="Generated Audio"),
css=custom_css
)
# ๋ฉ์ธ ์คํ ๋ถ๋ถ ์์
if __name__ == "__main__":
gr.TabbedInterface(
[video_search_tab, video_to_audio_tab, text_to_audio_tab],
["Video Search", "Video-to-Audio", "Text-to-Audio"],
css=custom_css
).launch(allowed_paths=[output_dir])
|