uxoxo/eb2ab
0
1import os2 3from lib.conf import tts_dir, voices_dir4loaded_tts = {}5 6TTS_ENGINES = {7 "XTTSv2": "xtts", 8 "BARK": "bark", 9 "VITS": "vits", 10 "FAIRSEQ": "fairseq", 11 "TACOTRON2": "tacotron", 12 "YOURTTS": "yourtts"13}14 15TTS_VOICE_CONVERSION = {16 "freevc24": {"path": "voice_conversion_models/multilingual/vctk/freevc24", "samplerate": 24000},17 "knnvc": {"path": "voice_conversion_models/multilingual/multi-dataset/knnvc", "samplerate": 16000},18 "openvoice_v1": {"path": "voice_conversion_models/multilingual/multi-dataset/openvoice_v1", "samplerate": 22050},19 "openvoice_v2": {"path": "voice_conversion_models/multilingual/multi-dataset/openvoice_v2", "samplerate": 22050}20}21 22TTS_SML = {23 "break": "‡break‡",24 "pause": "‡pause‡",25 "###": "‡pause‡"26}27 28default_tts_engine = TTS_ENGINES['XTTSv2']29default_fine_tuned = 'internal'30default_vc_model = TTS_VOICE_CONVERSION['knnvc']['path']31default_voice_detection_model = 'drewThomasson/segmentation'32 33max_tts_in_memory = 2 # TTS engines to keep in memory (1 tts engine ~= 4GB to 8GB RAM).34max_custom_model = 10035max_custom_voices = 100036max_upload_size = '6GB'37 38default_engine_settings = {39 TTS_ENGINES['XTTSv2']: {40 "samplerate": 24000,41 "temperature": 0.75,42 "length_penalty": 1.0,43 "num_beams": 1,44 "repetition_penalty": 3.0,45 "top_k": 50,46 "top_p": 0.85,47 "speed": 1.0,48 "enable_text_splitting": False,49 # to enable deepspeed, you must install it first:50 # conda activate ./python_env (linux/mac) or .\python_env (windows)51 # pip install deepspeed52 # conda deactivate53 "use_deepspeed": False,54 "files": ['config.json', 'model.pth', 'vocab.json', 'ref.wav', 'speakers_xtts.pth'],55 "voices": {56 "ClaribelDervla": "Claribel Dervla", "DaisyStudious": "Daisy Studious", "GracieWise": "Gracie Wise",57 "TammieEma": "Tammie Ema", "AlisonDietlinde": "Alison Dietlinde", "AnaFlorence": "Ana Florence",58 "AnnmarieNele": "Annmarie Nele", "AsyaAnara": "Asya Anara", "BrendaStern": "Brenda Stern",59 "GittaNikolina": "Gitta Nikolina", "HenrietteUsha": "Henriette Usha", "SofiaHellen": "Sofia Hellen",60 "TammyGrit": "Tammy Grit", "TanjaAdelina": "Tanja Adelina", "VjollcaJohnnie": "Vjollca Johnnie",61 "AndrewChipper": "Andrew Chipper", "BadrOdhiambo": "Badr Odhiambo", "DionisioSchuyler": "Dionisio Schuyler",62 "RoystonMin": "Royston Min", "ViktorEka": "Viktor Eka", "AbrahanMack": "Abrahan Mack",63 "AddeMichal": "Adde Michal", "BaldurSanjin": "Baldur Sanjin", "CraigGutsy": "Craig Gutsy",64 "DamienBlack": "Damien Black", "GilbertoMathias": "Gilberto Mathias", "IlkinUrbano": "Ilkin Urbano",65 "KazuhikoAtallah": "Kazuhiko Atallah", "LudvigMilivoj": "Ludvig Milivoj", "SuadQasim": "Suad Qasim",66 "TorcullDiarmuid": "Torcull Diarmuid", "ViktorMenelaos": "Viktor Menelaos", "ZacharieAimilios": "Zacharie Aimilios",67 "NovaHogarth": "Nova Hogarth", "MajaRuoho": "Maja Ruoho", "UtaObando": "Uta Obando",68 "LidiyaSzekeres": "Lidiya Szekeres", "ChandraMacFarland": "Chandra MacFarland", "SzofiGranger": "Szofi Granger",69 "CamillaHolmström": "Camilla Holmström", "LilyaStainthorpe": "Lilya Stainthorpe", "ZofijaKendrick": "Zofija Kendrick",70 "NarelleMoon": "Narelle Moon", "BarboraMacLean": "Barbora MacLean", "AlexandraHisakawa": "Alexandra Hisakawa",71 "AlmaMaría": "Alma María", "RosemaryOkafor": "Rosemary Okafor", "IgeBehringer": "Ige Behringer",72 "FilipTraverse": "Filip Traverse", "DamjanChapman": "Damjan Chapman", "WulfCarlevaro": "Wulf Carlevaro",73 "AaronDreschner": "Aaron Dreschner", "KumarDahl": "Kumar Dahl", "EugenioMataracı": "Eugenio Mataracı",74 "FerranSimen": "Ferran Simen", "XavierHayasaka": "Xavier Hayasaka", "LuisMoray": "Luis Moray",75 "MarcosRudaski": "Marcos Rudaski"76 },77 "rating": {"GPU VRAM": 4, "CPU": 3, "RAM": 8, "Realism": 4}78 },79 TTS_ENGINES['BARK']: {80 "samplerate": 24000,81 "text_temp": 0.50,82 "waveform_temp": 0.50,83 "files": ["text_2.pt", "coarse_2.pt", "fine_2.pt"],84 "speakers_path": os.path.join(voices_dir, '__bark'),85 "voices": {86 "de_speaker_0": "Speaker 0", "de_speaker_1": "Speaker 1", "de_speaker_2": "Speaker 2",87 "de_speaker_3": "Speaker 3", "de_speaker_4": "Speaker 4", "de_speaker_5": "Speaker 5",88 "de_speaker_6": "Speaker 6", "de_speaker_7": "Speaker 7", "de_speaker_8": "Speaker 8",89 "de_speaker_9": "Speaker 9", "en_speaker_0": "Speaker 0", "en_speaker_1": "Speaker 1",90 "en_speaker_2": "Speaker 2", "en_speaker_3": "Speaker 3", "en_speaker_4": "Speaker 4",91 "en_speaker_5": "Speaker 5", "en_speaker_6": "Speaker 6", "en_speaker_7": "Speaker 7",92 "en_speaker_8": "Speaker 8", "en_speaker_9": "Speaker 9", "es_speaker_0": "Speaker 0",93 "es_speaker_1": "Speaker 1", "es_speaker_2": "Speaker 2", "es_speaker_3": "Speaker 3",94 "es_speaker_4": "Speaker 4", "es_speaker_5": "Speaker 5", "es_speaker_6": "Speaker 6",95 "es_speaker_7": "Speaker 7", "es_speaker_8": "Speaker 8", "es_speaker_9": "Speaker 9",96 "fr_speaker_0": "Speaker 0", "fr_speaker_1": "Speaker 1", "fr_speaker_2": "Speaker 2",97 "fr_speaker_3": "Speaker 3", "fr_speaker_4": "Speaker 4", "fr_speaker_5": "Speaker 5",98 "fr_speaker_6": "Speaker 6", "fr_speaker_7": "Speaker 7", "fr_speaker_8": "Speaker 8",99 "fr_speaker_9": "Speaker 9", "hi_speaker_0": "Speaker 0", "hi_speaker_1": "Speaker 1",100 "hi_speaker_2": "Speaker 2", "hi_speaker_3": "Speaker 3", "hi_speaker_4": "Speaker 4",101 "hi_speaker_5": "Speaker 5", "hi_speaker_6": "Speaker 6", "hi_speaker_7": "Speaker 7",102 "hi_speaker_8": "Speaker 8", "hi_speaker_9": "Speaker 9", "it_speaker_0": "Speaker 0",103 "it_speaker_1": "Speaker 1", "it_speaker_2": "Speaker 2", "it_speaker_3": "Speaker 3",104 "it_speaker_4": "Speaker 4", "it_speaker_5": "Speaker 5", "it_speaker_6": "Speaker 6",105 "it_speaker_7": "Speaker 7", "it_speaker_8": "Speaker 8", "it_speaker_9": "Speaker 9",106 "ja_speaker_0": "Speaker 0", "ja_speaker_1": "Speaker 1", "ja_speaker_2": "Speaker 2",107 "ja_speaker_3": "Speaker 3", "ja_speaker_4": "Speaker 4", "ja_speaker_5": "Speaker 5",108 "ja_speaker_6": "Speaker 6", "ja_speaker_7": "Speaker 7", "ja_speaker_8": "Speaker 8",109 "ja_speaker_9": "Speaker 9", "ko_speaker_0": "Speaker 0", "ko_speaker_1": "Speaker 1",110 "ko_speaker_2": "Speaker 2", "ko_speaker_3": "Speaker 3", "ko_speaker_4": "Speaker 4",111 "ko_speaker_5": "Speaker 5", "ko_speaker_6": "Speaker 6", "ko_speaker_7": "Speaker 7",112 "ko_speaker_8": "Speaker 8", "ko_speaker_9": "Speaker 9", "pl_speaker_0": "Speaker 0",113 "pl_speaker_1": "Speaker 1", "pl_speaker_2": "Speaker 2", "pl_speaker_3": "Speaker 3",114 "pl_speaker_4": "Speaker 4", "pl_speaker_5": "Speaker 5", "pl_speaker_6": "Speaker 6",115 "pl_speaker_7": "Speaker 7", "pl_speaker_8": "Speaker 8", "pl_speaker_9": "Speaker 9",116 "pt_speaker_0": "Speaker 0", "pt_speaker_1": "Speaker 1", "pt_speaker_2": "Speaker 2",117 "pt_speaker_3": "Speaker 3", "pt_speaker_4": "Speaker 4", "pt_speaker_5": "Speaker 5",118 "pt_speaker_6": "Speaker 6", "pt_speaker_7": "Speaker 7", "pt_speaker_8": "Speaker 8",119 "pt_speaker_9": "Speaker 9", "ru_speaker_0": "Speaker 0", "ru_speaker_1": "Speaker 1",120 "ru_speaker_2": "Speaker 2", "ru_speaker_3": "Speaker 3", "ru_speaker_4": "Speaker 4",121 "ru_speaker_5": "Speaker 5", "ru_speaker_6": "Speaker 6", "ru_speaker_7": "Speaker 7",122 "ru_speaker_8": "Speaker 8", "ru_speaker_9": "Speaker 9", "tr_speaker_0": "Speaker 0",123 "tr_speaker_1": "Speaker 1", "tr_speaker_2": "Speaker 2", "tr_speaker_3": "Speaker 3",124 "tr_speaker_4": "Speaker 4", "tr_speaker_5": "Speaker 5", "tr_speaker_6": "Speaker 6",125 "tr_speaker_7": "Speaker 7", "tr_speaker_8": "Speaker 8", "tr_speaker_9": "Speaker 9",126 "zh_speaker_0": "Speaker 0", "zh_speaker_1": "Speaker 1", "zh_speaker_2": "Speaker 2",127 "zh_speaker_3": "Speaker 3", "zh_speaker_4": "Speaker 4", "zh_speaker_5": "Speaker 5",128 "zh_speaker_6": "Speaker 6", "zh_speaker_7": "Speaker 7", "zh_speaker_8": "Speaker 8",129 "zh_speaker_9": "Speaker 9"130 },131 "rating": {"GPU VRAM": 4, "CPU": 1, "RAM": 16, "Realism": 3}132 },133 TTS_ENGINES['VITS']: {134 "samplerate": 22050,135 "files": ['config.json', 'model_file.pth', 'language_ids.json'],136 "voices": {},137 "rating": {"GPU VRAM": 2, "CPU": 3, "RAM": 4, "Realism": 2}138 },139 TTS_ENGINES['FAIRSEQ']: {140 "samplerate": 16000,141 "files": ['config.json', 'G_100000.pth', 'vocab.json'],142 "voices": {},143 "rating": {"GPU VRAM": 2, "CPU": 3, "RAM": 4, "Realism": 2}144 },145 TTS_ENGINES['TACOTRON2']: {146 "samplerate": 22050,147 "files": ['config.json', 'best_model.pth', 'vocoder_config.json', 'vocoder_model.pth'],148 "voices": {},149 "rating": {"GPU VRAM": 2, "CPU": 3, "RAM": 4, "Realism": 2}150 },151 TTS_ENGINES['YOURTTS']: {152 "samplerate": 16000,153 "files": ['config.json', 'model_file.pth'],154 "voices": {"Machinella-5": "female-en-5", "ElectroMale-2": "male-en-2", 'Machinella-4': 'female-pt-4\n', 'ElectroMale-3': 'male-pt-3\n'},155 "rating": {"GPU VRAM": 1, "CPU": 5, "RAM": 4, "Realism": 1}156 }157}158models = {159 TTS_ENGINES['XTTSv2']: {160 "internal": {161 "lang": "multi",162 "repo": "coqui/XTTS-v2",163 "sub": "tts_models/multilingual/multi-dataset/xtts_v2/",164 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'KumarDahl.wav'),165 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],166 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']167 },168 "AiExplained": {169 "lang": "eng",170 "repo": "drewThomasson/fineTunedTTSModels",171 "sub": "xtts-v2/eng/AiExplained/",172 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'AiExplained.wav'),173 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],174 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']175 },176 "AsmrRacoon": {177 "lang": "eng",178 "repo": "drewThomasson/fineTunedTTSModels",179 "sub": "xtts-v2/eng/AsmrRacoon/",180 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'AsmrRacoon.wav'),181 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],182 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']183 },184 "Awkwafina": {185 "lang": "eng",186 "repo": "drewThomasson/fineTunedTTSModels",187 "sub": "xtts-v2/eng/Awkwafina/",188 "voice": os.path.join(voices_dir, 'eng', 'adult', 'female', 'Awkwafina.wav'),189 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],190 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']191 },192 "BobOdenkirk": {193 "lang": "eng",194 "repo": "drewThomasson/fineTunedTTSModels",195 "sub": "xtts-v2/eng/BobOdenkirk/",196 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'BobOdenkirk.wav'),197 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],198 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']199 },200 "BobRoss": {201 "lang": "eng",202 "repo": "drewThomasson/fineTunedTTSModels",203 "sub": "xtts-v2/eng/BobRoss/",204 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'BobRoss.wav'),205 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],206 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']207 },208 "BrinaPalencia": {209 "lang": "eng",210 "repo": "drewThomasson/fineTunedTTSModels",211 "sub": "xtts-v2/eng/BrinaPalencia/",212 "voice": os.path.join(voices_dir, 'eng', 'adult', 'female', 'BrinaPalencia.wav'),213 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],214 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']215 },216 "BryanCranston": {217 "lang": "eng",218 "repo": "drewThomasson/fineTunedTTSModels",219 "sub": "xtts-v2/eng/BryanCranston/",220 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'BryanCranston.wav'),221 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],222 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']223 },224 "DavidAttenborough": {225 "lang": "eng",226 "repo": "drewThomasson/fineTunedTTSModels",227 "sub": "xtts-v2/eng/DavidAttenborough/",228 "voice": os.path.join(voices_dir, 'eng', 'elder', 'male', 'DavidAttenborough.wav'),229 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],230 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']231 },232 "DeathPussInBoots": {233 "lang": "eng",234 "repo": "drewThomasson/fineTunedTTSModels",235 "sub": "xtts-v2/eng/DeathPussInBoots/",236 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'DeathPussInBoots.wav'),237 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],238 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']239 },240 "DermotCrowley": {241 "lang": "eng",242 "repo": "drewThomasson/fineTunedTTSModels",243 "sub": "xtts-v2/eng/DermotCrowley/",244 "voice": os.path.join(voices_dir, 'eng', 'elder', 'male', 'DermotCrowley.wav'),245 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],246 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']247 },248 "EvaSeymour": {249 "lang": "eng",250 "repo": "drewThomasson/fineTunedTTSModels",251 "sub": "xtts-v2/eng/EvaSeymour/",252 "voice": os.path.join(voices_dir, 'eng', 'adult', 'female', 'EvaSeymour.wav'),253 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],254 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']255 },256 "GideonOfnirEldenRing": {257 "lang": "eng",258 "repo": "drewThomasson/fineTunedTTSModels",259 "sub": "xtts-v2/eng/GideonOfnirEldenRing/",260 "voice": os.path.join(voices_dir, 'eng', 'elder', 'male', 'GideonOfnirEldenRing.wav'),261 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],262 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']263 },264 "GhostMW2": {265 "lang": "eng",266 "repo": "drewThomasson/fineTunedTTSModels",267 "sub": "xtts-v2/eng/GhostMW2/",268 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'GhostMW2.wav'),269 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],270 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']271 },272 "JhonButlerASMR": {273 "lang": "eng",274 "repo": "drewThomasson/fineTunedTTSModels",275 "sub": "xtts-v2/eng/JhonButlerASMR/",276 "voice": os.path.join(voices_dir, 'eng', 'elder', 'male', 'JhonButlerASMR.wav'),277 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],278 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']279 },280 "JhonMulaney": {281 "lang": "eng",282 "repo": "drewThomasson/fineTunedTTSModels",283 "sub": "xtts-v2/eng/JhonMulaney/",284 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'JhonMulaney.wav'),285 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],286 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']287 },288 "JillRedfield": {289 "lang": "eng",290 "repo": "drewThomasson/fineTunedTTSModels",291 "sub": "xtts-v2/eng/JillRedfield/",292 "voice": os.path.join(voices_dir, 'eng', 'adult', 'female', 'JillRedfield.wav'),293 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],294 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']295 },296 "JuliaWhenlan": {297 "lang": "eng",298 "repo": "drewThomasson/fineTunedTTSModels",299 "sub": "xtts-v2/eng/JuliaWhenlan/",300 "voice": os.path.join(voices_dir, 'eng', 'adult', 'female', 'JuliaWhenlan.wav'),301 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],302 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']303 },304 "LeeHorsley": {305 "lang": "eng",306 "repo": "drewThomasson/fineTunedTTSModels",307 "sub": "xtts-v2/eng/LeeHorsley/",308 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'LeeHorsley.wav'),309 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],310 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']311 },312 "MelinaEldenRing": {313 "lang": "eng",314 "repo": "drewThomasson/fineTunedTTSModels",315 "sub": "xtts-v2/eng/MelinaEldenRing/",316 "voice": os.path.join(voices_dir, 'eng', 'adult', 'female', 'MelinaEldenRing.wav'),317 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],318 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']319 },320 "MorganFreeman": {321 "lang": "eng",322 "repo": "drewThomasson/fineTunedTTSModels",323 "sub": "xtts-v2/eng/MorganFreeman/",324 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'MorganFreeman.wav'),325 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],326 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']327 },328 "NeilGaiman": {329 "lang": "eng",330 "repo": "drewThomasson/fineTunedTTSModels",331 "sub": "xtts-v2/eng/NeilGaiman/",332 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'NeilGaiman.wav'),333 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],334 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']335 },336 "RainyDayHeadSpace": {337 "lang": "eng",338 "repo": "drewThomasson/fineTunedTTSModels",339 "sub": "xtts-v2/eng/RainyDayHeadSpace/",340 "voice": os.path.join(voices_dir, 'eng', 'elder', 'male', 'RainyDayHeadSpace.wav'),341 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],342 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']343 },344 "RayPorter": {345 "lang": "eng",346 "repo": "drewThomasson/fineTunedTTSModels",347 "sub": "xtts-v2/eng/RayPorter/",348 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'RayPorter.wav'),349 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],350 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']351 },352 "RelaxForAWhile": {353 "lang": "eng",354 "repo": "drewThomasson/fineTunedTTSModels",355 "sub": "xtts-v2/eng/RelaxForAWhile/",356 "voice": os.path.join(voices_dir, 'eng', 'adult', 'female', 'RelaxForAWhile.wav'),357 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],358 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']359 },360 "RosamundPike": {361 "lang": "eng",362 "repo": "drewThomasson/fineTunedTTSModels",363 "sub": "xtts-v2/eng/RosamundPike/",364 "voice": os.path.join(voices_dir, 'eng', 'adult', 'female', 'RosamundPike.wav'),365 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],366 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']367 },368 "ScarlettJohansson": {369 "lang": "eng",370 "repo": "drewThomasson/fineTunedTTSModels",371 "sub": "xtts-v2/eng/ScarlettJohansson/",372 "voice": os.path.join(voices_dir, 'eng', 'adult', 'female', 'ScarlettJohansson.wav'),373 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],374 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']375 },376 "SladeTeenTitans": {377 "lang": "eng",378 "repo": "drewThomasson/fineTunedTTSModels",379 "sub": "xtts-v2/eng/SladeTeenTitans/",380 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'SladeTeenTitans.wav'),381 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],382 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']383 },384 "StanleyParable": {385 "lang": "eng",386 "repo": "drewThomasson/fineTunedTTSModels",387 "sub": "xtts-v2/eng/StanleyParable/",388 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'StanleyParable.wav'),389 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],390 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']391 },392 "Top15s": {393 "lang": "eng",394 "repo": "drewThomasson/fineTunedTTSModels",395 "sub": "xtts-v2/eng/Top15s/",396 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'Top15s.wav'),397 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],398 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']399 },400 "WhisperSalemASMR": {401 "lang": "eng",402 "repo": "drewThomasson/fineTunedTTSModels",403 "sub": "xtts-v2/eng/WhisperSalemASMR/",404 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'WhisperSalemASMR.wav'),405 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],406 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']407 },408 "Konishev": {409 "lang": "rus",410 "repo": "drewThomasson/fineTunedTTSModels",411 "sub": "xtts-v2/rus/Konishev/",412 "voice": os.path.join(voices_dir, 'rus', 'adult', 'male', 'Konishev.wav'),413 "files": default_engine_settings[TTS_ENGINES['XTTSv2']]['files'],414 "samplerate": default_engine_settings[TTS_ENGINES['XTTSv2']]['samplerate']415 }416 },417 TTS_ENGINES['BARK']: {418 "internal": {419 "lang": "multi",420 "repo": "erogol/bark", # suno/bark, rsxdalv/suno, tts_models/multilingual/multi-dataset/bark421 "sub": "", # {"big-bf16": "big-bf16/", "small-bf16": "small-bf16/", "big": "big/", "small": "small/"}422 "voice": os.path.join(voices_dir, 'eng', 'adult', 'male', 'KumarDahl.wav'),423 "files": default_engine_settings[TTS_ENGINES['BARK']]['files'],424 "samplerate": default_engine_settings[TTS_ENGINES['BARK']]['samplerate']425 }426 },427 TTS_ENGINES['VITS']: {428 "internal": {429 "lang": "multi",430 "repo": "tts_models/[lang_iso1]/[xxx]",431 "sub": {432 "css10/vits": ['es','hu','fi','fr','nl','ru','el'],433 "custom/vits": ['ca'],434 "custom/vits-female": ['bn', 'fa'],435 "cv/vits": ['bg','cs','da','et','ga','hr','lt','lv','mt','pt','ro','sk','sl','sv'],436 "mai/vits": ['uk'],437 "mai_female/vits": ['pl'],438 "mai_male/vits": ['it'],439 "openbible/vits": ['ewe','hau','lin','tw_akuapem','tw_asante','yor'],440 "vctk/vits": ['en'],441 "thorsten/vits": ['de']442 },443 "voice": None,444 "files": default_engine_settings[TTS_ENGINES['VITS']]['files'],445 "samplerate": {446 "css10/vits": default_engine_settings[TTS_ENGINES['VITS']]['samplerate'],447 "custom/vits": default_engine_settings[TTS_ENGINES['VITS']]['samplerate'],448 "custom/vits-female": default_engine_settings[TTS_ENGINES['VITS']]['samplerate'],449 "cv/vits": default_engine_settings[TTS_ENGINES['VITS']]['samplerate'],450 "mai/vits": default_engine_settings[TTS_ENGINES['VITS']]['samplerate'],451 "mai_female/vits": 24000,452 "mai_male/vits": 16000,453 "openbible/vits": default_engine_settings[TTS_ENGINES['VITS']]['samplerate'],454 "vctk/vits": default_engine_settings[TTS_ENGINES['VITS']]['samplerate'],455 "thorsten/vits": default_engine_settings[TTS_ENGINES['VITS']]['samplerate']456 }457 }458 },459 TTS_ENGINES['FAIRSEQ']: {460 "internal": {461 "lang": "multi",462 "repo": "tts_models/[lang]/fairseq/vits",463 "sub": "",464 "voice": None,465 "files": default_engine_settings[TTS_ENGINES['FAIRSEQ']]['files'],466 "samplerate": default_engine_settings[TTS_ENGINES['FAIRSEQ']]['samplerate']467 }468 },469 TTS_ENGINES['TACOTRON2']: {470 "internal": {471 "lang": "multi",472 "repo": "tts_models/[lang_iso1]/[xxx]",473 "sub": {474 "mai/tacotron2-DDC": ['fr', 'es', 'nl'],475 "thorsten/tacotron2-DDC": ['de'],476 "kokoro/tacotron2-DDC": ['ja'],477 "ljspeech/tacotron2-DDC": ['en'],478 "baker/tacotron2-DDC-GST": ['zh-CN'] 479 },480 "voice": None,481 "files": default_engine_settings[TTS_ENGINES['TACOTRON2']]['files'],482 "samplerate": {483 "mai/tacotron2-DDC": default_engine_settings[TTS_ENGINES['TACOTRON2']]['samplerate'],484 "thorsten/tacotron2-DDC": default_engine_settings[TTS_ENGINES['TACOTRON2']]['samplerate'],485 "kokoro/tacotron2-DDC": default_engine_settings[TTS_ENGINES['TACOTRON2']]['samplerate'],486 "ljspeech/tacotron2-DDC": default_engine_settings[TTS_ENGINES['TACOTRON2']]['samplerate'],487 "baker/tacotron2-DDC-GST": default_engine_settings[TTS_ENGINES['TACOTRON2']]['samplerate']488 },489 }490 },491 TTS_ENGINES['YOURTTS']: {492 "internal": {493 "lang": "multi",494 "repo": "tts_models/multilingual/multi-dataset/your_tts",495 "sub": "",496 "voice": None,497 "files": default_engine_settings[TTS_ENGINES['YOURTTS']]['files'],498 "samplerate": default_engine_settings[TTS_ENGINES['YOURTTS']]['samplerate']499 }500 }501}502 503 