Spaces:
Running
Running
File size: 26,805 Bytes
0a9bde8 9eae167 7b061c8 0a9bde8 9eae167 0a9bde8 7b061c8 0a9bde8 875b30b 0a9bde8 875b30b 0a9bde8 9eae167 0a9bde8 9eae167 0a9bde8 875b30b 0a9bde8 875b30b 0a9bde8 875b30b 4246b22 0a9bde8 875b30b 0a9bde8 875b30b 0a9bde8 875b30b 0a9bde8 875b30b 9eae167 875b30b 9eae167 875b30b 0a9bde8 875b30b 9eae167 875b30b 0a9bde8 875b30b 0a9bde8 875b30b 9eae167 875b30b 0a9bde8 875b30b 0a9bde8 875b30b 9eae167 875b30b 9eae167 875b30b 9eae167 875b30b 0a9bde8 875b30b 9eae167 875b30b 0a9bde8 875b30b 9eae167 0a9bde8 875b30b 9eae167 875b30b 0a9bde8 875b30b 0a9bde8 875b30b 9eae167 0a9bde8 875b30b 0a9bde8 875b30b 0a9bde8 875b30b 0a9bde8 533b363 7b061c8 |
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 |
import gradio as gr
import os
import argparse
from modules.whisper_Inference import WhisperInference
from modules.faster_whisper_inference import FasterWhisperInference
from modules.nllb_inference import NLLBInference
from ui.htmls import *
from modules.youtube_manager import get_ytmetas
from modules.deepl_api import DeepLAPI
from modules.whisper_parameter import *
# Ensure the outputs directory exists
def ensure_output_directory():
output_directories = ["outputs", "outputs/translations"]
for directory in output_directories:
if not os.path.exists(directory):
os.makedirs(directory)
class App:
def __init__(self, args):
ensure_output_directory() # Making sure the output directories exist
self.args = args
self.app = gr.Blocks(css=CSS, theme=self.args.theme)
self.whisper_inf = self.init_whisper()
print(f"Use \"{self.args.whisper_type}\" implementation")
print(f"Device \"{self.whisper_inf.device}\" is detected")
self.nllb_inf = NLLBInference()
self.deepl_api = DeepLAPI()
def init_whisper(self):
whisper_type = self.args.whisper_type.lower().strip()
if whisper_type in ["faster_whisper", "faster-whisper"]:
whisper_inf = FasterWhisperInference()
whisper_inf.model_dir = self.args.faster_whisper_model_dir
elif whisper_type == "whisper":
whisper_inf = WhisperInference()
whisper_inf.model_dir = self.args.whisper_model_dir
else:
# Default to FasterWhisperInference
whisper_inf = FasterWhisperInference()
whisper_inf.model_dir = self.args.faster_whisper_model_dir
return whisper_inf
@staticmethod
def open_folder(folder_path: str):
if os.path.exists(folder_path):
os.system(f"start {folder_path}")
else:
print(f"The folder {folder_path} does not exist.")
@staticmethod
def on_change_models(model_size: str):
translatable_model = ["large", "large-v1", "large-v2", "large-v3"]
if model_size not in translatable_model:
return gr.Checkbox(visible=False, value=False, interactive=False)
else:
return gr.Checkbox(visible=True, value=False, label="Translate to English?", interactive=True)
def launch(self):
with self.app:
with gr.Row():
with gr.Column():
gr.Markdown(MARKDOWN, elem_id="md_project")
with gr.Tabs():
with gr.TabItem("File"): # tab1
with gr.Row():
input_file = gr.Files(type="filepath", label="Upload File here")
with gr.Row():
dd_model = gr.Dropdown(choices=self.whisper_inf.available_models, value="large-v2",
label="Model")
dd_lang = gr.Dropdown(choices=["Automatic Detection"] + self.whisper_inf.available_langs,
value="Automatic Detection", label="Language")
dd_file_format = gr.Dropdown(["SRT", "WebVTT", "txt"], value="SRT", label="File Format")
with gr.Row():
cb_translate = gr.Checkbox(value=False, label="Translate to English?", interactive=True)
with gr.Row():
cb_timestamp = gr.Checkbox(value=True, label="Add a timestamp to the end of the filename", interactive=True)
with gr.Accordion("VAD Options", open=False, visible=isinstance(self.whisper_inf, FasterWhisperInference)):
cb_vad_filter = gr.Checkbox(label="Enable Silero VAD Filter", value=False, interactive=True)
sd_threshold = gr.Slider(minimum=0.0, maximum=1.0, step=0.01, label="Speech Threshold", value=0.5)
nb_min_speech_duration_ms = gr.Number(label="Minimum Speech Duration (ms)", precision=0, value=250)
nb_max_speech_duration_s = gr.Number(label="Maximum Speech Duration (s)", value=9999)
nb_min_silence_duration_ms = gr.Number(label="Minimum Silence Duration (ms)", precision=0, value=2000)
nb_window_size_sample = gr.Number(label="Window Size (samples)", precision=0, value=1024)
nb_speech_pad_ms = gr.Number(label="Speech Padding (ms)", precision=0, value=400)
with gr.Accordion("Advanced_Parameters", open=False):
nb_beam_size = gr.Number(label="Beam Size", value=1, precision=0, interactive=True)
nb_log_prob_threshold = gr.Number(label="Log Probability Threshold", value=-1.0, interactive=True)
nb_no_speech_threshold = gr.Number(label="No Speech Threshold", value=0.6, interactive=True)
dd_compute_type = gr.Dropdown(label="Compute Type", choices=self.whisper_inf.available_compute_types, value=self.whisper_inf.current_compute_type, interactive=True)
nb_best_of = gr.Number(label="Best Of", value=5, interactive=True)
nb_patience = gr.Number(label="Patience", value=1, interactive=True)
cb_condition_on_previous_text = gr.Checkbox(label="Condition On Previous Text", value=True, interactive=True)
tb_initial_prompt = gr.Textbox(label="Initial Prompt", value=None, interactive=True)
sd_temperature = gr.Slider(label="Temperature", value=0, step=0.01, maximum=1.0, interactive=True)
nb_compression_ratio_threshold = gr.Number(label="Compression Ratio Threshold", value=2.4, interactive=True)
with gr.Row():
btn_run = gr.Button("GENERATE SUBTITLE FILE", variant="primary")
with gr.Row():
tb_indicator = gr.Textbox(label="Output", scale=5)
files_subtitles = gr.Files(label="Downloadable output file", scale=3, interactive=False)
btn_openfolder = gr.Button('📂', scale=1)
params = [input_file, dd_file_format, cb_timestamp]
whisper_params = WhisperGradioComponents(model_size=dd_model,
lang=dd_lang,
is_translate=cb_translate,
beam_size=nb_beam_size,
log_prob_threshold=nb_log_prob_threshold,
no_speech_threshold=nb_no_speech_threshold,
compute_type=dd_compute_type,
best_of=nb_best_of,
patience=nb_patience,
condition_on_previous_text=cb_condition_on_previous_text,
initial_prompt=tb_initial_prompt,
temperature=sd_temperature,
compression_ratio_threshold=nb_compression_ratio_threshold,
vad_filter=cb_vad_filter,
threshold=sd_threshold,
min_speech_duration_ms=nb_min_speech_duration_ms,
max_speech_duration_s=nb_max_speech_duration_s,
min_silence_duration_ms=nb_min_silence_duration_ms,
window_size_sample=nb_window_size_sample,
speech_pad_ms=nb_speech_pad_ms)
btn_run.click(fn=self.whisper_inf.transcribe_file,
inputs=params + whisper_params.to_list(),
outputs=[tb_indicator, files_subtitles])
btn_openfolder.click(fn=lambda: self.open_folder("outputs"), inputs=None, outputs=None)
dd_model.change(fn=self.on_change_models, inputs=[dd_model], outputs=[cb_translate])
with gr.TabItem("Youtube"): # tab2
with gr.Row():
tb_youtubelink = gr.Textbox(label="Youtube Link")
with gr.Row(equal_height=True):
with gr.Column():
img_thumbnail = gr.Image(label="Youtube Thumbnail")
with gr.Column():
tb_title = gr.Label(label="Youtube Title")
tb_description = gr.Textbox(label="Youtube Description", max_lines=15)
with gr.Row():
dd_model = gr.Dropdown(choices=self.whisper_inf.available_models, value="large-v2", label="Model")
dd_lang = gr.Dropdown(choices=["Automatic Detection"] + self.whisper_inf.available_langs,
value="Automatic Detection", label="Language")
dd_file_format = gr.Dropdown(choices=["SRT", "WebVTT", "txt"], value="SRT", label="File Format")
with gr.Row():
cb_translate = gr.Checkbox(value=False, label="Translate to English?", interactive=True)
with gr.Row():
cb_timestamp = gr.Checkbox(value=True, label="Add a timestamp to the end of the filename", interactive=True)
with gr.Accordion("VAD Options", open=False, visible=isinstance(self.whisper_inf, FasterWhisperInference)):
cb_vad_filter = gr.Checkbox(label="Enable Silero VAD Filter", value=False, interactive=True)
sd_threshold = gr.Slider(minimum=0.0, maximum=1.0, step=0.01, label="Speech Threshold", value=0.5)
nb_min_speech_duration_ms = gr.Number(label="Minimum Speech Duration (ms)", precision=0, value=250)
nb_max_speech_duration_s = gr.Number(label="Maximum Speech Duration (s)", value=9999)
nb_min_silence_duration_ms = gr.Number(label="Minimum Silence Duration (ms)", precision=0, value=2000)
nb_window_size_sample = gr.Number(label="Window Size (samples)", precision=0, value=1024)
nb_speech_pad_ms = gr.Number(label="Speech Padding (ms)", precision=0, value=400)
with gr.Accordion("Advanced_Parameters", open=False):
nb_beam_size = gr.Number(label="Beam Size", value=1, precision=0, interactive=True)
nb_log_prob_threshold = gr.Number(label="Log Probability Threshold", value=-1.0, interactive=True)
nb_no_speech_threshold = gr.Number(label="No Speech Threshold", value=0.6, interactive=True)
dd_compute_type = gr.Dropdown(label="Compute Type", choices=self.whisper_inf.available_compute_types, value=self.whisper_inf.current_compute_type, interactive=True)
nb_best_of = gr.Number(label="Best Of", value=5, interactive=True)
nb_patience = gr.Number(label="Patience", value=1, interactive=True)
cb_condition_on_previous_text = gr.Checkbox(label="Condition On Previous Text", value=True, interactive=True)
tb_initial_prompt = gr.Textbox(label="Initial Prompt", value=None, interactive=True)
sd_temperature = gr.Slider(label="Temperature", value=0, step=0.01, maximum=1.0, interactive=True)
nb_compression_ratio_threshold = gr.Number(label="Compression Ratio Threshold", value=2.4, interactive=True)
with gr.Row():
btn_run = gr.Button("GENERATE SUBTITLE FILE", variant="primary")
with gr.Row():
tb_indicator = gr.Textbox(label="Output", scale=5)
files_subtitles = gr.Files(label="Downloadable output file", scale=3)
btn_openfolder = gr.Button('📂', scale=1)
params = [tb_youtubelink, dd_file_format, cb_timestamp]
whisper_params = WhisperGradioComponents(model_size=dd_model,
lang=dd_lang,
is_translate=cb_translate,
beam_size=nb_beam_size,
log_prob_threshold=nb_log_prob_threshold,
no_speech_threshold=nb_no_speech_threshold,
compute_type=dd_compute_type,
best_of=nb_best_of,
patience=nb_patience,
condition_on_previous_text=cb_condition_on_previous_text,
initial_prompt=tb_initial_prompt,
temperature=sd_temperature,
compression_ratio_threshold=nb_compression_ratio_threshold,
vad_filter=cb_vad_filter,
threshold=sd_threshold,
min_speech_duration_ms=nb_min_speech_duration_ms,
max_speech_duration_s=nb_max_speech_duration_s,
min_silence_duration_ms=nb_min_silence_duration_ms,
window_size_sample=nb_window_size_sample,
speech_pad_ms=nb_speech_pad_ms)
btn_run.click(fn=self.whisper_inf.transcribe_youtube,
inputs=params + whisper_params.to_list(),
outputs=[tb_indicator, files_subtitles])
tb_youtubelink.change(get_ytmetas, inputs=[tb_youtubelink],
outputs=[img_thumbnail, tb_title, tb_description])
btn_openfolder.click(fn=lambda: self.open_folder("outputs"), inputs=None, outputs=None)
dd_model.change(fn=self.on_change_models, inputs=[dd_model], outputs=[cb_translate])
with gr.TabItem("Mic"): # tab3
with gr.Row():
mic_input = gr.Microphone(label="Record with Mic", type="filepath", interactive=True)
with gr.Row():
dd_model = gr.Dropdown(choices=self.whisper_inf.available_models, value="large-v2",
label="Model")
dd_lang = gr.Dropdown(choices=["Automatic Detection"] + self.whisper_inf.available_langs,
value="Automatic Detection", label="Language")
dd_file_format = gr.Dropdown(["SRT", "WebVTT", "txt"], value="SRT", label="File Format")
with gr.Row():
cb_translate = gr.Checkbox(value=False, label="Translate to English?", interactive=True)
with gr.Accordion("VAD Options", open=False, visible=isinstance(self.whisper_inf, FasterWhisperInference)):
cb_vad_filter = gr.Checkbox(label="Enable Silero VAD Filter", value=False, interactive=True)
sd_threshold = gr.Slider(minimum=0.0, maximum=1.0, step=0.01, label="Speech Threshold", value=0.5)
nb_min_speech_duration_ms = gr.Number(label="Minimum Speech Duration (ms)", precision=0, value=250)
nb_max_speech_duration_s = gr.Number(label="Maximum Speech Duration (s)", value=9999)
nb_min_silence_duration_ms = gr.Number(label="Minimum Silence Duration (ms)", precision=0, value=2000)
nb_window_size_sample = gr.Number(label="Window Size (samples)", precision=0, value=1024)
nb_speech_pad_ms = gr.Number(label="Speech Padding (ms)", precision=0, value=400)
with gr.Accordion("Advanced_Parameters", open=False):
nb_beam_size = gr.Number(label="Beam Size", value=1, precision=0, interactive=True)
nb_log_prob_threshold = gr.Number(label="Log Probability Threshold", value=-1.0, interactive=True)
nb_no_speech_threshold = gr.Number(label="No Speech Threshold", value=0.6, interactive=True)
dd_compute_type = gr.Dropdown(label="Compute Type", choices=self.whisper_inf.available_compute_types, value=self.whisper_inf.current_compute_type, interactive=True)
nb_best_of = gr.Number(label="Best Of", value=5, interactive=True)
nb_patience = gr.Number(label="Patience", value=1, interactive=True)
cb_condition_on_previous_text = gr.Checkbox(label="Condition On Previous Text", value=True, interactive=True)
tb_initial_prompt = gr.Textbox(label="Initial Prompt", value=None, interactive=True)
sd_temperature = gr.Slider(label="Temperature", value=0, step=0.01, maximum=1.0, interactive=True)
with gr.Row():
btn_run = gr.Button("GENERATE SUBTITLE FILE", variant="primary")
with gr.Row():
tb_indicator = gr.Textbox(label="Output", scale=5)
files_subtitles = gr.Files(label="Downloadable output file", scale=3)
btn_openfolder = gr.Button('📂', scale=1)
params = [mic_input, dd_file_format]
whisper_params = WhisperGradioComponents(model_size=dd_model,
lang=dd_lang,
is_translate=cb_translate,
beam_size=nb_beam_size,
log_prob_threshold=nb_log_prob_threshold,
no_speech_threshold=nb_no_speech_threshold,
compute_type=dd_compute_type,
best_of=nb_best_of,
patience=nb_patience,
condition_on_previous_text=cb_condition_on_previous_text,
initial_prompt=tb_initial_prompt,
temperature=sd_temperature,
compression_ratio_threshold=nb_compression_ratio_threshold,
vad_filter=cb_vad_filter,
threshold=sd_threshold,
min_speech_duration_ms=nb_min_speech_duration_ms,
max_speech_duration_s=nb_max_speech_duration_s,
min_silence_duration_ms=nb_min_silence_duration_ms,
window_size_sample=nb_window_size_sample,
speech_pad_ms=nb_speech_pad_ms)
btn_run.click(fn=self.whisper_inf.transcribe_mic,
inputs=params + whisper_params.to_list(),
outputs=[tb_indicator, files_subtitles])
btn_openfolder.click(fn=lambda: self.open_folder("outputs"), inputs=None, outputs=None)
dd_model.change(fn=self.on_change_models, inputs=[dd_model], outputs=[cb_translate])
with gr.TabItem("T2T Translation"): # tab 4
with gr.Row():
file_subs = gr.Files(type="filepath", label="Upload Subtitle Files to translate here",
file_types=['.vtt', '.srt'])
with gr.TabItem("DeepL API"): # sub tab1
with gr.Row():
tb_authkey = gr.Textbox(label="Your Auth Key (API KEY)", value="")
with gr.Row():
dd_deepl_sourcelang = gr.Dropdown(label="Source Language", value="Automatic Detection",
choices=list(self.deepl_api.available_source_langs.keys()))
dd_deepl_targetlang = gr.Dropdown(label="Target Language", value="English",
choices=list(self.deepl_api.available_target_langs.keys()))
with gr.Row():
cb_deepl_ispro = gr.Checkbox(label="Pro User?", value=False)
with gr.Row():
btn_run = gr.Button("TRANSLATE SUBTITLE FILE", variant="primary")
with gr.Row():
tb_indicator = gr.Textbox(label="Output", scale=5)
files_subtitles = gr.Files(label="Downloadable output file", scale=3)
btn_openfolder = gr.Button('📂', scale=1)
btn_run.click(fn=self.deepl_api.translate_deepl,
inputs=[tb_authkey, file_subs, dd_deepl_sourcelang, dd_deepl_targetlang, cb_deepl_ispro],
outputs=[tb_indicator, files_subtitles])
btn_openfolder.click(fn=lambda: self.open_folder(os.path.join("outputs", "translations")),
inputs=None, outputs=None)
with gr.TabItem("NLLB"): # sub tab2
with gr.Row():
dd_nllb_model = gr.Dropdown(label="Model", value="facebook/nllb-200-1.3B",
choices=self.nllb_inf.available_models)
dd_nllb_sourcelang = gr.Dropdown(label="Source Language",
choices=self.nllb_inf.available_source_langs)
dd_nllb_targetlang = gr.Dropdown(label="Target Language", choices=self.nllb_inf.available_target_langs)
with gr.Row():
cb_timestamp = gr.Checkbox(value=True, label="Add a timestamp to the end of the filename", interactive=True)
with gr.Row():
btn_run = gr.Button("TRANSLATE SUBTITLE FILE", variant="primary")
with gr.Row():
tb_indicator = gr.Textbox(label="Output", scale=5)
files_subtitles = gr.Files(label="Downloadable output file", scale=3)
btn_openfolder = gr.Button('📂', scale=1)
with gr.Column():
md_vram_table = gr.HTML(NLLB_VRAM_TABLE, elem_id="md_nllb_vram_table")
btn_run.click(fn=self.nllb_inf.translate_file,
inputs=[file_subs, dd_nllb_model, dd_nllb_sourcelang, dd_nllb_targetlang, cb_timestamp],
outputs=[tb_indicator, files_subtitles])
btn_openfolder.click(fn=lambda: self.open_folder(os.path.join("outputs", "translations")),
inputs=None, outputs=None)
# Launch the app with optional gradio settings
launch_args = {}
if self.args.share:
launch_args['share'] = self.args.share
if self.args.server_name:
launch_args['server_name'] = self.args.server_name
if self.args.server_port:
launch_args['server_port'] = self.args.server_port
if self.args.username and self.args.password:
launch_args['auth'] = (self.args.username, self.args.password)
launch_args['inbrowser'] = True
self.app.queue(api_open=False).launch(**launch_args)
# Create the parser for command-line arguments
parser = argparse.ArgumentParser()
parser.add_argument('--whisper_type', type=str, default="faster-whisper", help='A type of the whisper implementation between: ["whisper", "faster-whisper"]')
parser.add_argument('--share', type=bool, default=False, nargs='?', const=True, help='Gradio share value')
parser.add_argument('--server_name', type=str, default=None, help='Gradio server host')
parser.add_argument('--server_port', type=int, default=None, help='Gradio server port')
parser.add_argument('--username', type=str, default=None, help='Gradio authentication username')
parser.add_argument('--password', type=str, default=None, help='Gradio authentication password')
parser.add_argument('--theme', type=str, default=None, help='Gradio Blocks theme')
parser.add_argument('--colab', type=bool, default=False, nargs='?', const=True, help='Is colab user or not')
parser.add_argument('--api_open', type=bool, default=False, nargs='?', const=True, help='enable api or not')
parser.add_argument('--whisper_model_dir', type=str, default=os.path.join("models", "Whisper"), help='Directory path of the whisper model')
parser.add_argument('--faster_whisper_model_dir', type=str, default=os.path.join("models", "Whisper", "faster-whisper"), help='Directory path of the faster-whisper model')
_args = parser.parse_args()
if __name__ == "__main__":
app = App(args=_args)
app.launch() |