Pendrokar commited on
Commit
e5a3534
Β·
1 Parent(s): a318e59

HTML title comments on models

Browse files
Files changed (1) hide show
  1. app/models.py +56 -6
app/models.py CHANGED
@@ -149,6 +149,8 @@ HF_SPACES = {
149
  'text_param_index': 0,
150
  'return_audio_index': 0,
151
  'series': 'XTTS',
 
 
152
  },
153
 
154
  # WhisperSpeech
@@ -159,6 +161,7 @@ HF_SPACES = {
159
  'return_audio_index': 0,
160
  'series': 'WhisperSpeech',
161
  'emoji': '😷', # broken space
 
162
  },
163
 
164
  # OpenVoice (MyShell.ai)
@@ -169,6 +172,7 @@ HF_SPACES = {
169
  'return_audio_index': 1,
170
  'series': 'OpenVoice',
171
  'emoji': '😩', # old gradio
 
172
  },
173
  # OpenVoice v2 (MyShell.ai)
174
  'myshell-ai/OpenVoiceV2': {
@@ -178,6 +182,7 @@ HF_SPACES = {
178
  'return_audio_index': 1,
179
  'series': 'OpenVoice',
180
  'emoji': '😩', # old gradio
 
181
  },
182
 
183
  # MetaVoice
@@ -188,6 +193,7 @@ HF_SPACES = {
188
  'return_audio_index': 0,
189
  'series': 'MetaVoice',
190
  'emoji': '😷', # broken space
 
191
  },
192
 
193
  # xVASynth (CPU)
@@ -197,6 +203,7 @@ HF_SPACES = {
197
  'text_param_index': 0,
198
  'return_audio_index': 0,
199
  'series': 'xVASynth',
 
200
  },
201
  'Pendrokar/xVASynth-TTS/NoDeepMoji': {
202
  'name': 'xVASynth v3',
@@ -205,6 +212,7 @@ HF_SPACES = {
205
  'return_audio_index': 0,
206
  'series': 'xVASynth',
207
  'space_link': 'Pendrokar/xVASynth-TTS',
 
208
  },
209
 
210
  # CoquiTTS (CPU)
@@ -214,6 +222,7 @@ HF_SPACES = {
214
  'text_param_index': 0,
215
  'return_audio_index': 0,
216
  'series': 'CoquiTTS',
 
217
  },
218
 
219
  # HierSpeech_TTS
@@ -223,7 +232,8 @@ HF_SPACES = {
223
  'text_param_index': 0,
224
  'return_audio_index': 0,
225
  'series': 'HierSpeech++',
226
- 'emoji': 'πŸ˜’', # unemotional
 
227
  },
228
 
229
  # MeloTTS (MyShell.ai)
@@ -233,6 +243,8 @@ HF_SPACES = {
233
  'text_param_index': 'text',
234
  'return_audio_index': 0,
235
  'series': 'MeloTTS',
 
 
236
  },
237
 
238
  # Parler Mini
@@ -244,6 +256,7 @@ HF_SPACES = {
244
  'is_zero_gpu_space': True,
245
  'series': 'Parler',
246
  'emoji': '😷', # broken space
 
247
  },
248
  # Parler Large
249
  'parler-tts/parler_tts/large': {
@@ -253,6 +266,9 @@ HF_SPACES = {
253
  'return_audio_index': 0,
254
  'is_zero_gpu_space': True,
255
  'series': 'Parler',
 
 
 
256
  },
257
  # Parler Mini trained on Expresso dataset
258
  'parler-tts/parler-tts-expresso': {
@@ -263,6 +279,7 @@ HF_SPACES = {
263
  'is_zero_gpu_space': True,
264
  'series': 'Parler',
265
  # 'emoji': 'πŸ˜ƒ', # overlly jolly voice
 
266
  },
267
 
268
  # Parler Mini trained on Expresso dataset
@@ -273,6 +290,7 @@ HF_SPACES = {
273
  'return_audio_index': 1, # 0 is the reformatted text
274
  'is_zero_gpu_space': True,
275
  'series': 'Parler',
 
276
  },
277
 
278
  # Parler Mini trained on Expresso dataset, reformats description using Gemma 2b
@@ -283,6 +301,7 @@ HF_SPACES = {
283
  'return_audio_index': 1, # 0 is the reformatted text
284
  'is_zero_gpu_space': True,
285
  'series': 'Parler',
 
286
  },
287
 
288
  # Microsoft Edge TTS
@@ -293,7 +312,8 @@ HF_SPACES = {
293
  'return_audio_index': 0,
294
  'is_closed_source': True,
295
  'series': 'Edge TTS',
296
- 'emoji': '', # api disabled
 
297
  'space_link': 'innoai/Edge-TTS-Text-to-Speech', # API disabled
298
  },
299
 
@@ -305,6 +325,7 @@ HF_SPACES = {
305
  'return_audio_index': 0,
306
  'series': 'Fish Speech',
307
  'emoji': '😡', # redirects to OpenAudio
 
308
  },
309
 
310
  # OpenAudio S1 (Fish Audio)
@@ -315,6 +336,7 @@ HF_SPACES = {
315
  'return_audio_index': 0,
316
  'series': 'Fish Speech',
317
  # 'emoji': '😷',
 
318
  },
319
 
320
  # F5 TTS
@@ -326,6 +348,7 @@ HF_SPACES = {
326
  'is_zero_gpu_space': True,
327
  # 'series': 'E2 TTS',
328
  'series': 'E2/F5 TTS',
 
329
  },
330
 
331
  # E2 TTS TODO: call switch model function
@@ -346,6 +369,7 @@ HF_SPACES = {
346
  'text_param_index': 0,
347
  'return_audio_index': 0,
348
  'series': 'IMS-Toucan',
 
349
  },
350
 
351
  # IMS-Toucan English non-artificial
@@ -355,7 +379,8 @@ HF_SPACES = {
355
  'text_param_index': 0,
356
  'return_audio_index': 0,
357
  'series': 'IMS-Toucan',
358
- 'emoji': 'πŸ˜’', # unemotional
 
359
  },
360
 
361
  # StyleTTS v2
@@ -366,7 +391,8 @@ HF_SPACES = {
366
  'return_audio_index': 0,
367
  'is_zero_gpu_space': True,
368
  'series': 'StyleTTS',
369
- # 'emoji': 'πŸ˜ͺ',
 
370
  },
371
 
372
  # StyleTTS Kokoro v0.19
@@ -378,6 +404,7 @@ HF_SPACES = {
378
  'is_zero_gpu_space': False,
379
  'series': 'Kokoro',
380
  'space_link': 'Remsky/Kokoro-TTS-Zero', # still supports v0.19
 
381
  },
382
 
383
  # StyleTTS Kokoro v0.23
@@ -388,6 +415,7 @@ HF_SPACES = {
388
  'return_audio_index': 0,
389
  'is_zero_gpu_space': True,
390
  'series': 'Kokoro',
 
391
  },
392
 
393
  # StyleTTS Kokoro v1.0 (CPU)
@@ -411,6 +439,7 @@ HF_SPACES = {
411
  'is_zero_gpu_space': True,
412
  'series': 'MaskGCT',
413
  'emoji': 'πŸ₯΅', # requires 300s reserved ZeroGPU!
 
414
  },
415
  'Svngoku/maskgct-audio-lab': {
416
  'name': 'MaskGCT',
@@ -420,6 +449,7 @@ HF_SPACES = {
420
  'is_zero_gpu_space': True,
421
  'series': 'MaskGCT',
422
  'emoji': 'πŸ₯΅', # requires 300s reserved ZeroGPU!
 
423
  },
424
 
425
  # GPT-SoVITS v2
@@ -430,6 +460,7 @@ HF_SPACES = {
430
  'return_audio_index': 0,
431
  'is_zero_gpu_space': True,
432
  'series': 'GPT-SoVITS',
 
433
  },
434
 
435
  # OuteTTS v0.2 500M
@@ -441,6 +472,7 @@ HF_SPACES = {
441
  'is_zero_gpu_space': True,
442
  'series': 'OuteTTS',
443
  'emoji': 'πŸ₯΅', # requires 300s reserved ZeroGPU!
 
444
  },
445
  # OuteTTS v0.3 1B
446
  'OuteAI/OuteTTS-0.3-1B-Demo': {
@@ -451,6 +483,7 @@ HF_SPACES = {
451
  'is_zero_gpu_space': True,
452
  'series': 'OuteTTS',
453
  'emoji': 'πŸ₯΅', # requires 300s reserved ZeroGPU!
 
454
  },
455
 
456
  # LlaSa 1B
@@ -462,6 +495,7 @@ HF_SPACES = {
462
  'is_zero_gpu_space': True,
463
  'series': 'LLaSA',
464
  # 'emoji': '😷', # broken space
 
465
  },
466
 
467
  # LlaSa 3B
@@ -473,6 +507,7 @@ HF_SPACES = {
473
  'is_zero_gpu_space': True,
474
  'series': 'LLaSA',
475
  # 'emoji': '😷', # broken space
 
476
  },
477
 
478
  # LlaSa 8B
@@ -483,7 +518,8 @@ HF_SPACES = {
483
  'return_audio_index': 0,
484
  'is_zero_gpu_space': True,
485
  'series': 'LLaSA',
486
- 'emoji': '😷', # broken space
 
487
  },
488
 
489
  # Mars5
@@ -505,6 +541,7 @@ HF_SPACES = {
505
  'is_zero_gpu_space': False,
506
  'is_closed_source': True,
507
  'series': 'MARS',
 
508
  },
509
 
510
  # Zonos
@@ -515,6 +552,7 @@ HF_SPACES = {
515
  'return_audio_index': 0,
516
  'is_zero_gpu_space': True,
517
  'series': 'Zonos',
 
518
  },
519
  'Steveeeeeeen/Zonos/hybrid': {
520
  'name': 'Zonos H',
@@ -523,6 +561,7 @@ HF_SPACES = {
523
  'return_audio_index': 0,
524
  'is_zero_gpu_space': True,
525
  'series': 'Zonos',
 
526
  },
527
 
528
  # Spark-TTS
@@ -533,6 +572,7 @@ HF_SPACES = {
533
  'return_audio_index': 0,
534
  'is_zero_gpu_space': True,
535
  'series': 'Spark-TTS',
 
536
  },
537
 
538
  'sesame/csm-1b' : {
@@ -542,6 +582,7 @@ HF_SPACES = {
542
  'return_audio_index': 0,
543
  'is_zero_gpu_space': True,
544
  'series': 'CSM-1B',
 
545
  },
546
 
547
  'MohamedRashad/Orpheus-TTS' : {
@@ -560,6 +601,7 @@ HF_SPACES = {
560
  'return_audio_index': 0,
561
  'is_zero_gpu_space': True,
562
  'series': 'Index',
 
563
  },
564
 
565
  'nari-labs/Dia-1.6B' : {
@@ -569,6 +611,7 @@ HF_SPACES = {
569
  'return_audio_index': 0,
570
  'is_zero_gpu_space': True,
571
  'series': 'Dia',
 
572
  },
573
 
574
  'ResembleAI/Chatterbox' : {
@@ -1028,7 +1071,14 @@ def make_link_to_space(model_name, for_leaderboard=False):
1028
  title += model_name
1029
  else:
1030
  style += 'font-style: italic;'
1031
- title += model_name +'; Disabled (See AVAILABLE_MODELS within code for why)'
 
 
 
 
 
 
 
1032
 
1033
  # bolden top five models which get more scrutinized
1034
  if model_name in top_five:
 
149
  'text_param_index': 0,
150
  'return_audio_index': 0,
151
  'series': 'XTTS',
152
+ 'emoji': '😩', # old gradio
153
+ 'title': 'Gradio version too old', # old gradio
154
  },
155
 
156
  # WhisperSpeech
 
161
  'return_audio_index': 0,
162
  'series': 'WhisperSpeech',
163
  'emoji': '😷', # broken space
164
+ 'title': 'Broken space - runtime error',
165
  },
166
 
167
  # OpenVoice (MyShell.ai)
 
172
  'return_audio_index': 1,
173
  'series': 'OpenVoice',
174
  'emoji': '😩', # old gradio
175
+ 'title': 'Gradio version too old', # old gradio
176
  },
177
  # OpenVoice v2 (MyShell.ai)
178
  'myshell-ai/OpenVoiceV2': {
 
182
  'return_audio_index': 1,
183
  'series': 'OpenVoice',
184
  'emoji': '😩', # old gradio
185
+ 'title': 'Gradio version too old', # old gradio
186
  },
187
 
188
  # MetaVoice
 
193
  'return_audio_index': 0,
194
  'series': 'MetaVoice',
195
  'emoji': '😷', # broken space
196
+ 'title': 'Broken space - runtime error',
197
  },
198
 
199
  # xVASynth (CPU)
 
203
  'text_param_index': 0,
204
  'return_audio_index': 0,
205
  'series': 'xVASynth',
206
+ 'title': 'Outclassed',
207
  },
208
  'Pendrokar/xVASynth-TTS/NoDeepMoji': {
209
  'name': 'xVASynth v3',
 
212
  'return_audio_index': 0,
213
  'series': 'xVASynth',
214
  'space_link': 'Pendrokar/xVASynth-TTS',
215
+ 'title': 'Outclassed',
216
  },
217
 
218
  # CoquiTTS (CPU)
 
222
  'text_param_index': 0,
223
  'return_audio_index': 0,
224
  'series': 'CoquiTTS',
225
+ 'title': 'Outclassed',
226
  },
227
 
228
  # HierSpeech_TTS
 
232
  'text_param_index': 0,
233
  'return_audio_index': 0,
234
  'series': 'HierSpeech++',
235
+ 'emoji': 'πŸ˜’', # Narration voice
236
+ 'title': 'Narration voice',
237
  },
238
 
239
  # MeloTTS (MyShell.ai)
 
243
  'text_param_index': 'text',
244
  'return_audio_index': 0,
245
  'series': 'MeloTTS',
246
+ 'emoji': 'πŸ˜’', # Narration voice
247
+ 'title': 'Narration voice',
248
  },
249
 
250
  # Parler Mini
 
256
  'is_zero_gpu_space': True,
257
  'series': 'Parler',
258
  'emoji': '😷', # broken space
259
+ 'title': 'Broken space - runtime error',
260
  },
261
  # Parler Large
262
  'parler-tts/parler_tts/large': {
 
266
  'return_audio_index': 0,
267
  'is_zero_gpu_space': True,
268
  'series': 'Parler',
269
+ 'emoji': '😷', # broken space
270
+ 'title': 'Broken space - runtime error',
271
+ 'space_link': 'parler-tts/parler_tts',
272
  },
273
  # Parler Mini trained on Expresso dataset
274
  'parler-tts/parler-tts-expresso': {
 
279
  'is_zero_gpu_space': True,
280
  'series': 'Parler',
281
  # 'emoji': 'πŸ˜ƒ', # overlly jolly voice
282
+ 'title': 'Overlly jolly voice', # overlly jolly voice
283
  },
284
 
285
  # Parler Mini trained on Expresso dataset
 
290
  'return_audio_index': 1, # 0 is the reformatted text
291
  'is_zero_gpu_space': True,
292
  'series': 'Parler',
293
+ 'title': 'Unstable',
294
  },
295
 
296
  # Parler Mini trained on Expresso dataset, reformats description using Gemma 2b
 
301
  'return_audio_index': 1, # 0 is the reformatted text
302
  'is_zero_gpu_space': True,
303
  'series': 'Parler',
304
+ 'title': 'Unstable',
305
  },
306
 
307
  # Microsoft Edge TTS
 
312
  'return_audio_index': 0,
313
  'is_closed_source': True,
314
  'series': 'Edge TTS',
315
+ 'emoji': '', # Gradio API disabled
316
+ # 'title': 'API Disabled',
317
  'space_link': 'innoai/Edge-TTS-Text-to-Speech', # API disabled
318
  },
319
 
 
325
  'return_audio_index': 0,
326
  'series': 'Fish Speech',
327
  'emoji': '😡', # redirects to OpenAudio
328
+ 'title': 'Deprecated: Redirects to OpenAudio', # redirects to OpenAudio
329
  },
330
 
331
  # OpenAudio S1 (Fish Audio)
 
336
  'return_audio_index': 0,
337
  'series': 'Fish Speech',
338
  # 'emoji': '😷',
339
+ # 'title': '😷',
340
  },
341
 
342
  # F5 TTS
 
348
  'is_zero_gpu_space': True,
349
  # 'series': 'E2 TTS',
350
  'series': 'E2/F5 TTS',
351
+ 'title': 'Uses a voice sample from trained dataset',
352
  },
353
 
354
  # E2 TTS TODO: call switch model function
 
369
  'text_param_index': 0,
370
  'return_audio_index': 0,
371
  'series': 'IMS-Toucan',
372
+ 'title': 'Changes voice pitch on each request',
373
  },
374
 
375
  # IMS-Toucan English non-artificial
 
379
  'text_param_index': 0,
380
  'return_audio_index': 0,
381
  'series': 'IMS-Toucan',
382
+ 'emoji': 'πŸ˜’', # Narration voice
383
+ 'title': 'Narration voice',
384
  },
385
 
386
  # StyleTTS v2
 
391
  'return_audio_index': 0,
392
  'is_zero_gpu_space': True,
393
  'series': 'StyleTTS',
394
+ 'emoji': '😷', # broken space
395
+ 'title': 'Outclassed by Kokoro',
396
  },
397
 
398
  # StyleTTS Kokoro v0.19
 
404
  'is_zero_gpu_space': False,
405
  'series': 'Kokoro',
406
  'space_link': 'Remsky/Kokoro-TTS-Zero', # still supports v0.19
407
+ 'title': 'Deprecated: Kokoro v1.0+ improves pronunciation',
408
  },
409
 
410
  # StyleTTS Kokoro v0.23
 
415
  'return_audio_index': 0,
416
  'is_zero_gpu_space': True,
417
  'series': 'Kokoro',
418
+ 'title': 'Deprecated: Kokoro v1.0+ improves pronunciation',
419
  },
420
 
421
  # StyleTTS Kokoro v1.0 (CPU)
 
439
  'is_zero_gpu_space': True,
440
  'series': 'MaskGCT',
441
  'emoji': 'πŸ₯΅', # requires 300s reserved ZeroGPU!
442
+ 'title': 'Requires 300s reserved ZeroGPU time! Cannot afford!',
443
  },
444
  'Svngoku/maskgct-audio-lab': {
445
  'name': 'MaskGCT',
 
449
  'is_zero_gpu_space': True,
450
  'series': 'MaskGCT',
451
  'emoji': 'πŸ₯΅', # requires 300s reserved ZeroGPU!
452
+ 'title': 'Requires 300s reserved ZeroGPU time! Cannot afford!',
453
  },
454
 
455
  # GPT-SoVITS v2
 
460
  'return_audio_index': 0,
461
  'is_zero_gpu_space': True,
462
  'series': 'GPT-SoVITS',
463
+ 'title': 'Outclassed',
464
  },
465
 
466
  # OuteTTS v0.2 500M
 
472
  'is_zero_gpu_space': True,
473
  'series': 'OuteTTS',
474
  'emoji': 'πŸ₯΅', # requires 300s reserved ZeroGPU!
475
+ 'title': 'Requires 300s reserved ZeroGPU time! Cannot afford!',
476
  },
477
  # OuteTTS v0.3 1B
478
  'OuteAI/OuteTTS-0.3-1B-Demo': {
 
483
  'is_zero_gpu_space': True,
484
  'series': 'OuteTTS',
485
  'emoji': 'πŸ₯΅', # requires 300s reserved ZeroGPU!
486
+ 'title': 'Requires 300s reserved ZeroGPU time! Cannot afford!',
487
  },
488
 
489
  # LlaSa 1B
 
495
  'is_zero_gpu_space': True,
496
  'series': 'LLaSA',
497
  # 'emoji': '😷', # broken space
498
+ 'title': 'Broken space - Runtime error',
499
  },
500
 
501
  # LlaSa 3B
 
507
  'is_zero_gpu_space': True,
508
  'series': 'LLaSA',
509
  # 'emoji': '😷', # broken space
510
+ # 'title': 'Broken space - ZeroGPU Pro account expired',
511
  },
512
 
513
  # LlaSa 8B
 
518
  'return_audio_index': 0,
519
  'is_zero_gpu_space': True,
520
  'series': 'LLaSA',
521
+ # 'emoji': '😷', # broken space
522
+ # 'title': 'Broken space - ZeroGPU Pro account expired',
523
  },
524
 
525
  # Mars5
 
541
  'is_zero_gpu_space': False,
542
  'is_closed_source': True,
543
  'series': 'MARS',
544
+ 'title': 'Unstable',
545
  },
546
 
547
  # Zonos
 
552
  'return_audio_index': 0,
553
  'is_zero_gpu_space': True,
554
  'series': 'Zonos',
555
+ 'title': 'Outclassed',
556
  },
557
  'Steveeeeeeen/Zonos/hybrid': {
558
  'name': 'Zonos H',
 
561
  'return_audio_index': 0,
562
  'is_zero_gpu_space': True,
563
  'series': 'Zonos',
564
+ 'title': 'Outclassed',
565
  },
566
 
567
  # Spark-TTS
 
572
  'return_audio_index': 0,
573
  'is_zero_gpu_space': True,
574
  'series': 'Spark-TTS',
575
+ 'title': 'Outclassed',
576
  },
577
 
578
  'sesame/csm-1b' : {
 
582
  'return_audio_index': 0,
583
  'is_zero_gpu_space': True,
584
  'series': 'CSM-1B',
585
+ 'title': 'Outclassed',
586
  },
587
 
588
  'MohamedRashad/Orpheus-TTS' : {
 
601
  'return_audio_index': 0,
602
  'is_zero_gpu_space': True,
603
  'series': 'Index',
604
+ 'title': 'Outclassed',
605
  },
606
 
607
  'nari-labs/Dia-1.6B' : {
 
611
  'return_audio_index': 0,
612
  'is_zero_gpu_space': True,
613
  'series': 'Dia',
614
+ 'title': 'Outclassed',
615
  },
616
 
617
  'ResembleAI/Chatterbox' : {
 
1071
  title += model_name
1072
  else:
1073
  style += 'font-style: italic;'
1074
+ title += model_name
1075
+
1076
+ # add HTML title with comment on model
1077
+ try:
1078
+ if HF_SPACES[model_name]['title']:
1079
+ title += '; '+ HF_SPACES[model_name]['title']
1080
+ except:
1081
+ pass
1082
 
1083
  # bolden top five models which get more scrutinized
1084
  if model_name in top_five: