Spaces:
Runtime error
Runtime error
File size: 12,438 Bytes
ee0444e 8fb8950 748b9ed 8fb8950 ecc5344 8fb8950 df28e5a c3b9be5 df28e5a c3b9be5 df28e5a 8fb8950 748b9ed ecc5344 748b9ed 4936337 ecc5344 8fb8950 df28e5a ecc5344 8fb8950 ecc5344 8fb8950 ecc5344 8fb8950 748b9ed df28e5a 748b9ed ecc5344 748b9ed df28e5a 8fb8950 ecc5344 748b9ed 4936337 748b9ed 4936337 748b9ed 4936337 748b9ed 4936337 748b9ed 4936337 748b9ed 4936337 748b9ed 4936337 748b9ed 8fb8950 748b9ed 8fb8950 748b9ed 9419e6a 748b9ed 8fb8950 748b9ed 8fb8950 748b9ed 9419e6a 748b9ed 4936337 748b9ed 8fb8950 748b9ed 9419e6a 748b9ed 4936337 748b9ed 4936337 748b9ed 4936337 748b9ed 4936337 8fb8950 748b9ed 8fb8950 748b9ed 8fb8950 748b9ed 9419e6a 748b9ed 8fb8950 748b9ed 9419e6a 748b9ed 4936337 748b9ed 4936337 8fb8950 748b9ed 8fb8950 29e6f40 748b9ed ecc5344 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 |
from __future__ import annotations
import os
import pathlib
import gradio as gr
import numpy as np
import torch
import torchaudio
from huggingface_hub import snapshot_download
from seamless_communication.inference import Translator
from lang_list import (
ASR_TARGET_LANGUAGE_NAMES,
LANGUAGE_NAME_TO_CODE,
S2ST_TARGET_LANGUAGE_NAMES,
S2TT_TARGET_LANGUAGE_NAMES,
T2ST_TARGET_LANGUAGE_NAMES,
T2TT_TARGET_LANGUAGE_NAMES,
TEXT_SOURCE_LANGUAGE_NAMES,
)
if not pathlib.Path("models").exists():
snapshot_download(repo_id="meta-private/M4Tv2", repo_type="model", local_dir="models")
DESCRIPTION = """\
# SeamlessM4T
[SeamlessM4T](https://github.com/facebookresearch/seamless_communication) is designed to provide high-quality
translation, allowing people from different linguistic communities to communicate effortlessly through speech and text.
This unified model enables multiple tasks like Speech-to-Speech (S2ST), Speech-to-Text (S2TT), Text-to-Speech (T2ST)
translation and more, without relying on multiple separate models.
"""
CACHE_EXAMPLES = os.getenv("CACHE_EXAMPLES") == "1" and torch.cuda.is_available()
AUDIO_SAMPLE_RATE = 16000.0
MAX_INPUT_AUDIO_LENGTH = 60 # in seconds
DEFAULT_TARGET_LANGUAGE = "French"
if torch.cuda.is_available():
device = torch.device("cuda:0")
dtype = torch.float16
else:
device = torch.device("cpu")
dtype = torch.float32
translator = Translator(
model_name_or_card="seamlessM4T_v2_large",
vocoder_name_or_card="vocoder_v2",
device=device,
dtype=dtype,
)
def preprocess_audio(input_audio: str) -> None:
arr, org_sr = torchaudio.load(input_audio)
new_arr = torchaudio.functional.resample(arr, orig_freq=org_sr, new_freq=AUDIO_SAMPLE_RATE)
max_length = int(MAX_INPUT_AUDIO_LENGTH * AUDIO_SAMPLE_RATE)
if new_arr.shape[1] > max_length:
new_arr = new_arr[:, :max_length]
gr.Warning(f"Input audio is too long. Only the first {MAX_INPUT_AUDIO_LENGTH} seconds is used.")
torchaudio.save(input_audio, new_arr, sample_rate=int(AUDIO_SAMPLE_RATE))
def run_s2st(input_audio: str, target_language: str) -> tuple[tuple[int, np.ndarray] | None, str]:
preprocess_audio(input_audio)
target_language_code = LANGUAGE_NAME_TO_CODE[target_language]
out_texts, out_audios = translator.predict(
input=input_audio,
task_str="S2ST",
tgt_lang=target_language_code,
)
out_text = str(out_texts[0])
out_wav = out_audios.audio_wavs[0].cpu().detach().numpy()
return (int(AUDIO_SAMPLE_RATE), out_wav), out_text
def run_s2tt(input_audio: str, target_language: str) -> str:
preprocess_audio(input_audio)
target_language_code = LANGUAGE_NAME_TO_CODE[target_language]
out_texts, _ = translator.predict(
input=input_audio,
task_str="S2TT",
tgt_lang=target_language_code,
)
return str(out_texts[0])
def run_t2st(input_text: str, source_language: str, target_language: str) -> tuple[tuple[int, np.ndarray] | None, str]:
source_language_code = LANGUAGE_NAME_TO_CODE[source_language]
target_language_code = LANGUAGE_NAME_TO_CODE[target_language]
out_texts, out_audios = translator.predict(
input=input_text,
task_str="T2ST",
tgt_lang=target_language_code,
src_lang=source_language_code,
)
out_text = str(out_texts[0])
out_wav = out_audios.audio_wavs[0].cpu().detach().numpy()
return (int(AUDIO_SAMPLE_RATE), out_wav), out_text
def run_t2tt(input_text: str, source_language: str, target_language: str) -> str:
source_language_code = LANGUAGE_NAME_TO_CODE[source_language]
target_language_code = LANGUAGE_NAME_TO_CODE[target_language]
out_texts, _ = translator.predict(
input=input_text,
task_str="T2TT",
tgt_lang=target_language_code,
src_lang=source_language_code,
)
return str(out_texts[0])
def run_asr(input_audio: str, target_language: str) -> str:
preprocess_audio(input_audio)
target_language_code = LANGUAGE_NAME_TO_CODE[target_language]
out_texts, _ = translator.predict(
input=input_audio,
task_str="ASR",
tgt_lang=target_language_code,
)
return str(out_texts[0])
with gr.Blocks() as demo_s2st:
with gr.Row():
with gr.Column():
with gr.Group():
target_language = gr.Dropdown(
label="Target language",
choices=S2ST_TARGET_LANGUAGE_NAMES,
value=DEFAULT_TARGET_LANGUAGE,
)
input_audio = gr.Audio(label="Input speech", type="filepath")
btn = gr.Button("Translate")
with gr.Column():
with gr.Group():
output_audio = gr.Audio(
label="Translated speech",
autoplay=False,
streaming=False,
type="numpy",
)
output_text = gr.Textbox(label="Translated text")
gr.Examples(
examples=[
["assets/sample_input.mp3", "French"],
["assets/sample_input.mp3", "Mandarin Chinese"],
["assets/sample_input_2.mp3", "Hindi"],
["assets/sample_input_2.mp3", "Spanish"],
],
inputs=[input_audio, target_language],
outputs=[output_audio, output_text],
fn=run_s2st,
cache_examples=CACHE_EXAMPLES,
api_name=False,
)
btn.click(
fn=run_s2st,
inputs=[input_audio, target_language],
outputs=[output_audio, output_text],
api_name="s2st",
)
with gr.Blocks() as demo_s2tt:
with gr.Row():
with gr.Column():
with gr.Group():
target_language = gr.Dropdown(
label="Target language",
choices=S2TT_TARGET_LANGUAGE_NAMES,
value=DEFAULT_TARGET_LANGUAGE,
)
input_audio = gr.Audio(label="Input speech", type="filepath")
btn = gr.Button("Translate")
with gr.Column():
output_text = gr.Textbox(label="Translated text")
gr.Examples(
examples=[
["assets/sample_input.mp3", "French"],
["assets/sample_input.mp3", "Mandarin Chinese"],
["assets/sample_input_2.mp3", "Hindi"],
["assets/sample_input_2.mp3", "Spanish"],
],
inputs=[input_audio, target_language],
outputs=output_text,
fn=run_s2tt,
cache_examples=CACHE_EXAMPLES,
api_name=False,
)
btn.click(
fn=run_s2tt,
inputs=[input_audio, target_language],
outputs=output_text,
api_name="s2tt",
)
with gr.Blocks() as demo_t2st:
with gr.Row():
with gr.Column():
with gr.Group():
with gr.Row():
source_language = gr.Dropdown(
label="Source language",
choices=TEXT_SOURCE_LANGUAGE_NAMES,
value="English",
)
target_language = gr.Dropdown(
label="Target language",
choices=T2ST_TARGET_LANGUAGE_NAMES,
value=DEFAULT_TARGET_LANGUAGE,
)
input_text = gr.Textbox(label="Input text")
btn = gr.Button("Translate")
with gr.Column():
with gr.Group():
output_audio = gr.Audio(
label="Translated speech",
autoplay=False,
streaming=False,
type="numpy",
)
output_text = gr.Textbox(label="Translated text")
gr.Examples(
examples=[
[
"My favorite animal is the elephant.",
"English",
"French",
],
[
"My favorite animal is the elephant.",
"English",
"Mandarin Chinese",
],
[
"Meta AI's Seamless M4T model is democratising spoken communication across language barriers",
"English",
"Hindi",
],
[
"Meta AI's Seamless M4T model is democratising spoken communication across language barriers",
"English",
"Spanish",
],
],
inputs=[input_text, source_language, target_language],
outputs=[output_audio, output_text],
fn=run_t2st,
cache_examples=CACHE_EXAMPLES,
api_name=False,
)
gr.on(
triggers=[input_text.submit, btn.click],
fn=run_t2st,
inputs=[input_text, source_language, target_language],
outputs=[output_audio, output_text],
api_name="t2st",
)
with gr.Blocks() as demo_t2tt:
with gr.Row():
with gr.Column():
with gr.Group():
with gr.Row():
source_language = gr.Dropdown(
label="Source language",
choices=TEXT_SOURCE_LANGUAGE_NAMES,
value="English",
)
target_language = gr.Dropdown(
label="Target language",
choices=T2TT_TARGET_LANGUAGE_NAMES,
value=DEFAULT_TARGET_LANGUAGE,
)
input_text = gr.Textbox(label="Input text")
btn = gr.Button("Translate")
with gr.Column():
output_text = gr.Textbox(label="Translated text")
gr.Examples(
examples=[
[
"My favorite animal is the elephant.",
"English",
"French",
],
[
"My favorite animal is the elephant.",
"English",
"Mandarin Chinese",
],
[
"Meta AI's Seamless M4T model is democratising spoken communication across language barriers",
"English",
"Hindi",
],
[
"Meta AI's Seamless M4T model is democratising spoken communication across language barriers",
"English",
"Spanish",
],
],
inputs=[input_text, source_language, target_language],
outputs=output_text,
fn=run_t2tt,
cache_examples=CACHE_EXAMPLES,
api_name=False,
)
gr.on(
triggers=[input_text.submit, btn.click],
fn=run_t2tt,
inputs=[input_text, source_language, target_language],
outputs=output_text,
api_name="t2tt",
)
with gr.Blocks() as demo_asr:
with gr.Row():
with gr.Column():
with gr.Group():
target_language = gr.Dropdown(
label="Target language",
choices=ASR_TARGET_LANGUAGE_NAMES,
value=DEFAULT_TARGET_LANGUAGE,
)
input_audio = gr.Audio(label="Input speech", type="filepath")
btn = gr.Button("Translate")
with gr.Column():
output_text = gr.Textbox(label="Translated text")
gr.Examples(
examples=[
["assets/sample_input.mp3", "English"],
["assets/sample_input_2.mp3", "English"],
],
inputs=[input_audio, target_language],
outputs=output_text,
fn=run_asr,
cache_examples=CACHE_EXAMPLES,
api_name=False,
)
btn.click(
fn=run_asr,
inputs=[input_audio, target_language],
outputs=output_text,
api_name="asr",
)
with gr.Blocks(css="style.css") as demo:
gr.Markdown(DESCRIPTION)
gr.DuplicateButton(
value="Duplicate Space for private use",
elem_id="duplicate-button",
visible=os.getenv("SHOW_DUPLICATE_BUTTON") == "1",
)
with gr.Tabs():
with gr.Tab(label="S2ST"):
demo_s2st.render()
with gr.Tab(label="S2TT"):
demo_s2tt.render()
with gr.Tab(label="T2ST"):
demo_t2st.render()
with gr.Tab(label="T2TT"):
demo_t2tt.render()
with gr.Tab(label="ASR"):
demo_asr.render()
if __name__ == "__main__":
demo.queue(max_size=50).launch()
|