@qvac/sdk 0.17.0 → 0.18.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (154) hide show
  1. package/NOTICE +83 -18
  2. package/README.md +19 -0
  3. package/dist/client/api/audio-gen.d.ts +17 -0
  4. package/dist/client/api/audio-gen.js +17 -0
  5. package/dist/client/api/finetune.d.ts +2 -0
  6. package/dist/client/api/finetune.js +7 -1
  7. package/dist/client/api/index.d.ts +1 -1
  8. package/dist/client/api/index.js +1 -1
  9. package/dist/client/api/load-model.d.ts +8 -0
  10. package/dist/client/api/transcribe.d.ts +19 -2
  11. package/dist/client/api/transcribe.js +66 -27
  12. package/dist/client/api/translate.d.ts +3 -2
  13. package/dist/client/api/translate.js +42 -21
  14. package/dist/client/api/vla.d.ts +42 -1
  15. package/dist/client/api/vla.js +50 -1
  16. package/dist/commands/bundle/entry-gen.d.ts +4 -0
  17. package/dist/commands/bundle/entry-gen.js +6 -0
  18. package/dist/commands/bundle/index.js +19 -12
  19. package/dist/examples/asr/parakeet-indic-conformer-filesystem.d.ts +2 -0
  20. package/dist/examples/asr/parakeet-indic-conformer-filesystem.js +52 -0
  21. package/dist/examples/audiogen/generate-cover.d.ts +2 -0
  22. package/dist/examples/audiogen/generate-cover.js +100 -0
  23. package/dist/examples/llamacpp-fallback.d.ts +2 -0
  24. package/dist/examples/llamacpp-fallback.js +34 -0
  25. package/dist/examples/multi-job-completion.d.ts +28 -0
  26. package/dist/examples/multi-job-completion.js +128 -0
  27. package/dist/examples/profiling/basic.js +32 -1
  28. package/dist/examples/system-resources.d.ts +2 -0
  29. package/dist/examples/system-resources.js +86 -0
  30. package/dist/examples/tts/audio8.d.ts +2 -0
  31. package/dist/examples/tts/audio8.js +54 -0
  32. package/dist/examples/tts/cosyvoice3.d.ts +2 -0
  33. package/dist/examples/tts/cosyvoice3.js +52 -0
  34. package/dist/examples/vla-groot.js +92 -49
  35. package/dist/index.d.ts +2 -2
  36. package/dist/index.js +2 -2
  37. package/dist/models/registry/models.d.ts +782 -6
  38. package/dist/models/registry/models.js +2664 -1859
  39. package/dist/schemas/audio-gen.d.ts +103 -0
  40. package/dist/schemas/audio-gen.js +135 -4
  41. package/dist/schemas/common.d.ts +550 -11
  42. package/dist/schemas/constants-registry.d.ts +18 -4
  43. package/dist/schemas/constants-registry.js +16 -3
  44. package/dist/schemas/index.d.ts +1 -1
  45. package/dist/schemas/index.js +1 -1
  46. package/dist/schemas/llamacpp-config.d.ts +10 -10
  47. package/dist/schemas/llamacpp-config.js +12 -5
  48. package/dist/schemas/load-model.d.ts +3678 -323
  49. package/dist/schemas/load-model.js +21 -3
  50. package/dist/schemas/sdk-config.d.ts +60 -24
  51. package/dist/schemas/sdk-errors-server.d.ts +1 -0
  52. package/dist/schemas/sdk-errors-server.js +5 -0
  53. package/dist/schemas/text-to-speech.d.ts +1086 -110
  54. package/dist/schemas/text-to-speech.js +194 -4
  55. package/dist/schemas/tools.d.ts +0 -16
  56. package/dist/schemas/tools.js +0 -15
  57. package/dist/schemas/transcription-config.d.ts +2 -0
  58. package/dist/schemas/transcription-config.js +10 -4
  59. package/dist/schemas/transcription.d.ts +3 -0
  60. package/dist/schemas/vla.d.ts +83 -0
  61. package/dist/schemas/vla.js +105 -2
  62. package/dist/scripts/link-workspace-inference.d.ts +2 -0
  63. package/dist/scripts/link-workspace-inference.js +31 -0
  64. package/dist/server/bare/ops/cancel.d.ts +1 -1
  65. package/dist/server/bare/ops/cancel.js +1 -1
  66. package/dist/server/bare/ops/kv-cache-retention.js +11 -0
  67. package/dist/server/bare/ops/kv-cache-utils.d.ts +1 -1
  68. package/dist/server/bare/ops/kv-cache-utils.js +40 -12
  69. package/dist/server/bare/ops/transcribe.d.ts +2 -2
  70. package/dist/server/bare/ops/transcribe.js +54 -48
  71. package/dist/server/bare/ops/translate-stats.d.ts +5 -0
  72. package/dist/server/bare/ops/translate-stats.js +11 -0
  73. package/dist/server/bare/ops/translate.js +66 -38
  74. package/dist/server/bare/ops/unload-model.js +30 -24
  75. package/dist/server/bare/plugins/audiogen-ggml/ops/audio-gen-input.d.ts +20 -0
  76. package/dist/server/bare/plugins/audiogen-ggml/ops/audio-gen-input.js +153 -0
  77. package/dist/server/bare/plugins/audiogen-ggml/ops/audio-gen-stream.js +52 -1
  78. package/dist/server/bare/plugins/audiogen-ggml/plugin.d.ts +28 -0
  79. package/dist/server/bare/plugins/ggml-ocr/plugin.js +3 -1
  80. package/dist/server/bare/plugins/ggml-vla/ops/vla-set-embodiment.d.ts +3 -0
  81. package/dist/server/bare/plugins/ggml-vla/ops/vla-set-embodiment.js +9 -0
  82. package/dist/server/bare/plugins/ggml-vla/plugin.d.ts +37 -0
  83. package/dist/server/bare/plugins/ggml-vla/plugin.js +21 -4
  84. package/dist/server/bare/plugins/llamacpp-completion/ops/batch-completion-stream.js +16 -15
  85. package/dist/server/bare/plugins/llamacpp-completion/ops/cache-logger.d.ts +1 -1
  86. package/dist/server/bare/plugins/llamacpp-completion/ops/cache-logger.js +2 -2
  87. package/dist/server/bare/plugins/llamacpp-completion/ops/completion-stats.d.ts +6 -0
  88. package/dist/server/bare/plugins/llamacpp-completion/ops/completion-stats.js +10 -0
  89. package/dist/server/bare/plugins/llamacpp-completion/ops/completion-stream.d.ts +1 -0
  90. package/dist/server/bare/plugins/llamacpp-completion/ops/completion-stream.js +131 -123
  91. package/dist/server/bare/plugins/llamacpp-completion/ops/finetune.js +110 -20
  92. package/dist/server/bare/plugins/llamacpp-completion/ops/kv-cache-session.d.ts +49 -23
  93. package/dist/server/bare/plugins/llamacpp-completion/ops/kv-cache-session.js +308 -143
  94. package/dist/server/bare/plugins/llamacpp-completion/ops/kv-cache-state.d.ts +1 -0
  95. package/dist/server/bare/plugins/llamacpp-completion/ops/kv-cache-state.js +2 -2
  96. package/dist/server/bare/plugins/llamacpp-completion/plugin.d.ts +6 -6
  97. package/dist/server/bare/plugins/llamacpp-completion/plugin.js +55 -36
  98. package/dist/server/bare/plugins/llamacpp-completion/transform.js +1 -7
  99. package/dist/server/bare/plugins/llamacpp-embedding/plugin.js +2 -3
  100. package/dist/server/bare/plugins/nmtcpp-translation/plugin.js +6 -5
  101. package/dist/server/bare/plugins/parakeet-transcription/plugin.js +27 -16
  102. package/dist/server/bare/plugins/tts-ggml/ops/parler-options.js +23 -3
  103. package/dist/server/bare/plugins/tts-ggml/plugin.d.ts +302 -4
  104. package/dist/server/bare/plugins/tts-ggml/plugin.js +131 -0
  105. package/dist/server/bare/plugins/whispercpp-transcription/plugin.js +48 -39
  106. package/dist/server/bare/registry/model-registry.d.ts +1 -0
  107. package/dist/server/bare/runtime/request-registry-singleton.js +29 -15
  108. package/dist/server/bare/runtime/request-registry.d.ts +27 -0
  109. package/dist/server/bare/runtime/request-registry.js +246 -76
  110. package/dist/server/bare/types/addon-responses.d.ts +4 -12
  111. package/dist/server/rpc/handlers/load-model/handler.js +2 -2
  112. package/dist/server/rpc/handlers/load-model/resolve-session.d.ts +1 -1
  113. package/dist/server/rpc/handlers/load-model/resolve-session.js +3 -3
  114. package/dist/server/rpc/handlers/load-model/resolve.d.ts +12 -2
  115. package/dist/server/rpc/handlers/load-model/resolve.js +90 -12
  116. package/dist/server/utils/audio/decoder.d.ts +37 -1
  117. package/dist/server/utils/audio/decoder.js +81 -51
  118. package/dist/server/utils/config-transform.d.ts +3 -0
  119. package/dist/server/utils/config-transform.js +6 -0
  120. package/dist/server/utils/index.d.ts +0 -1
  121. package/dist/server/utils/index.js +0 -1
  122. package/dist/server/utils/shard-utils.d.ts +1 -0
  123. package/dist/server/utils/shard-utils.js +14 -0
  124. package/dist/server/utils/tool-integration.d.ts +3 -10
  125. package/dist/server/utils/tool-integration.js +3 -12
  126. package/dist/utils/errors-server.d.ts +3 -0
  127. package/dist/utils/errors-server.js +5 -0
  128. package/package.json +12 -14
  129. package/dist/examples/llamacpp-dynamic-tools.d.ts +0 -2
  130. package/dist/examples/llamacpp-dynamic-tools.js +0 -152
  131. package/dist/models/update-models/codegen.d.ts +0 -3
  132. package/dist/models/update-models/codegen.js +0 -196
  133. package/dist/models/update-models/companions.d.ts +0 -30
  134. package/dist/models/update-models/companions.js +0 -296
  135. package/dist/models/update-models/history.d.ts +0 -24
  136. package/dist/models/update-models/history.js +0 -107
  137. package/dist/models/update-models/index.d.ts +0 -2
  138. package/dist/models/update-models/index.js +0 -123
  139. package/dist/models/update-models/naming.d.ts +0 -4
  140. package/dist/models/update-models/naming.js +0 -396
  141. package/dist/models/update-models/processing.d.ts +0 -8
  142. package/dist/models/update-models/processing.js +0 -65
  143. package/dist/models/update-models/registry.d.ts +0 -5
  144. package/dist/models/update-models/registry.js +0 -87
  145. package/dist/models/update-models/shards.d.ts +0 -4
  146. package/dist/models/update-models/shards.js +0 -57
  147. package/dist/models/update-models/types.d.ts +0 -88
  148. package/dist/models/update-models/types.js +0 -1
  149. package/dist/models/update-models/utils.d.ts +0 -3
  150. package/dist/models/update-models/utils.js +0 -21
  151. package/dist/server/bare/plugins/llamacpp-completion/ops/batch-cancelled.d.ts +0 -23
  152. package/dist/server/bare/plugins/llamacpp-completion/ops/batch-cancelled.js +0 -30
  153. package/dist/server/utils/expand-gguf-shards.d.ts +0 -3
  154. package/dist/server/utils/expand-gguf-shards.js +0 -22
@@ -2,6 +2,11 @@ import { z } from 'zod';
2
2
  export declare const TTS_CHATTERBOX_LANGUAGES: readonly ["en", "es", "fr", "de", "it", "ja", "pt", "nl", "pl", "tr", "sv", "da", "fi", "no", "el", "ms", "sw", "ar", "ko", "he", "ru", "zh", "hi"];
3
3
  export declare const TTS_SUPERTONIC_LANGUAGES: readonly ["en", "ko", "ja", "ar", "bg", "cs", "da", "de", "el", "es", "et", "fi", "fr", "hi", "hr", "hu", "id", "it", "lt", "lv", "nl", "pl", "pt", "ro", "ru", "sk", "sl", "sv", "tr", "uk", "vi"];
4
4
  export declare const TTS_PARLER_EMOTIONS: readonly ["command", "anger", "narration", "conversation", "disgust", "fear", "happy", "neutral", "proper noun", "news", "sad", "surprise"];
5
+ export declare const TTS_PACES: readonly ["slow", "moderate", "fast"];
6
+ export declare const TTS_COSYVOICE3_EMOTIONS: readonly ["anger", "happy", "neutral", "sad"];
7
+ export declare const TTS_COSYVOICE3_INSTRUCT_DIALECTS: readonly ["cantonese", "northeastern", "gansu", "guizhou", "henan", "hubei", "hunan", "jiangxi", "minnan", "ningxia", "shanxi", "shaanxi", "shandong", "shanghai", "sichuan", "tianjin", "yunnan"];
8
+ export declare const TTS_COSYVOICE3_INSTRUCT_VOLUMES: readonly ["loud", "soft"];
9
+ export declare const TTS_COSYVOICE3_INSTRUCT_STYLES: readonly ["peppa", "robot"];
5
10
  export declare const TTS_LANGUAGES: readonly ["en", "es", "fr", "de", "it", "ja", "pt", "nl", "pl", "tr", "sv", "da", "fi", "no", "el", "ms", "sw", "ar", "ko", "he", "ru", "zh", "hi", "bg", "cs", "et", "hr", "hu", "id", "lt", "lv", "ro", "sk", "sl", "uk", "vi"];
6
11
  export declare const ttsChatterboxRuntimeConfigSchema: z.ZodObject<{
7
12
  ttsEngine: z.ZodLiteral<"chatterbox">;
@@ -114,13 +119,110 @@ export declare const ttsParlerRuntimeConfigSchema: z.ZodObject<{
114
119
  surprise: "surprise";
115
120
  }>>;
116
121
  pitch: z.ZodOptional<z.ZodString>;
117
- pace: z.ZodOptional<z.ZodString>;
122
+ pace: z.ZodOptional<z.ZodEnum<{
123
+ slow: "slow";
124
+ moderate: "moderate";
125
+ fast: "fast";
126
+ }>>;
118
127
  expressivity: z.ZodOptional<z.ZodString>;
119
128
  noise: z.ZodOptional<z.ZodString>;
120
129
  reverb: z.ZodOptional<z.ZodString>;
121
130
  quality: z.ZodOptional<z.ZodString>;
122
131
  ttsEngine: z.ZodLiteral<"parler">;
123
132
  }, z.core.$strip>;
133
+ declare const ttsCosyvoice3InstructSchema: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
134
+ dialect: z.ZodOptional<z.ZodEnum<{
135
+ cantonese: "cantonese";
136
+ northeastern: "northeastern";
137
+ gansu: "gansu";
138
+ guizhou: "guizhou";
139
+ henan: "henan";
140
+ hubei: "hubei";
141
+ hunan: "hunan";
142
+ jiangxi: "jiangxi";
143
+ minnan: "minnan";
144
+ ningxia: "ningxia";
145
+ shanxi: "shanxi";
146
+ shaanxi: "shaanxi";
147
+ shandong: "shandong";
148
+ shanghai: "shanghai";
149
+ sichuan: "sichuan";
150
+ tianjin: "tianjin";
151
+ yunnan: "yunnan";
152
+ }>>;
153
+ volume: z.ZodOptional<z.ZodEnum<{
154
+ loud: "loud";
155
+ soft: "soft";
156
+ }>>;
157
+ style: z.ZodOptional<z.ZodEnum<{
158
+ peppa: "peppa";
159
+ robot: "robot";
160
+ }>>;
161
+ }, z.core.$strict>]>;
162
+ export declare const ttsCosyvoice3RuntimeConfigSchema: z.ZodObject<{
163
+ ttsEngine: z.ZodLiteral<"cosyvoice3">;
164
+ emotion: z.ZodOptional<z.ZodEnum<{
165
+ anger: "anger";
166
+ happy: "happy";
167
+ neutral: "neutral";
168
+ sad: "sad";
169
+ }>>;
170
+ pace: z.ZodOptional<z.ZodEnum<{
171
+ slow: "slow";
172
+ moderate: "moderate";
173
+ fast: "fast";
174
+ }>>;
175
+ instruct: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
176
+ dialect: z.ZodOptional<z.ZodEnum<{
177
+ cantonese: "cantonese";
178
+ northeastern: "northeastern";
179
+ gansu: "gansu";
180
+ guizhou: "guizhou";
181
+ henan: "henan";
182
+ hubei: "hubei";
183
+ hunan: "hunan";
184
+ jiangxi: "jiangxi";
185
+ minnan: "minnan";
186
+ ningxia: "ningxia";
187
+ shanxi: "shanxi";
188
+ shaanxi: "shaanxi";
189
+ shandong: "shandong";
190
+ shanghai: "shanghai";
191
+ sichuan: "sichuan";
192
+ tianjin: "tianjin";
193
+ yunnan: "yunnan";
194
+ }>>;
195
+ volume: z.ZodOptional<z.ZodEnum<{
196
+ loud: "loud";
197
+ soft: "soft";
198
+ }>>;
199
+ style: z.ZodOptional<z.ZodEnum<{
200
+ peppa: "peppa";
201
+ robot: "robot";
202
+ }>>;
203
+ }, z.core.$strict>]>>;
204
+ useGPU: z.ZodOptional<z.ZodBoolean>;
205
+ outputSampleRate: z.ZodOptional<z.ZodNumber>;
206
+ streamChunkTokens: z.ZodOptional<z.ZodNumber>;
207
+ streamFirstChunkTokens: z.ZodOptional<z.ZodNumber>;
208
+ threads: z.ZodOptional<z.ZodNumber>;
209
+ nGpuLayers: z.ZodOptional<z.ZodNumber>;
210
+ seed: z.ZodOptional<z.ZodNumber>;
211
+ }, z.core.$strip>;
212
+ export declare const ttsAudio8RuntimeConfigSchema: z.ZodObject<{
213
+ ttsEngine: z.ZodLiteral<"audio8">;
214
+ referenceText: z.ZodOptional<z.ZodString>;
215
+ greedy: z.ZodOptional<z.ZodBoolean>;
216
+ temperature: z.ZodOptional<z.ZodNumber>;
217
+ topK: z.ZodOptional<z.ZodNumber>;
218
+ topP: z.ZodOptional<z.ZodNumber>;
219
+ maxFrames: z.ZodOptional<z.ZodNumber>;
220
+ useGPU: z.ZodOptional<z.ZodBoolean>;
221
+ outputSampleRate: z.ZodOptional<z.ZodNumber>;
222
+ threads: z.ZodOptional<z.ZodNumber>;
223
+ nGpuLayers: z.ZodOptional<z.ZodNumber>;
224
+ seed: z.ZodOptional<z.ZodNumber>;
225
+ }, z.core.$strip>;
124
226
  export declare const ttsRuntimeConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
125
227
  ttsEngine: z.ZodLiteral<"chatterbox">;
126
228
  language: z.ZodEnum<{
@@ -230,12 +332,78 @@ export declare const ttsRuntimeConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObjec
230
332
  surprise: "surprise";
231
333
  }>>;
232
334
  pitch: z.ZodOptional<z.ZodString>;
233
- pace: z.ZodOptional<z.ZodString>;
335
+ pace: z.ZodOptional<z.ZodEnum<{
336
+ slow: "slow";
337
+ moderate: "moderate";
338
+ fast: "fast";
339
+ }>>;
234
340
  expressivity: z.ZodOptional<z.ZodString>;
235
341
  noise: z.ZodOptional<z.ZodString>;
236
342
  reverb: z.ZodOptional<z.ZodString>;
237
343
  quality: z.ZodOptional<z.ZodString>;
238
344
  ttsEngine: z.ZodLiteral<"parler">;
345
+ }, z.core.$strip>, z.ZodObject<{
346
+ ttsEngine: z.ZodLiteral<"cosyvoice3">;
347
+ emotion: z.ZodOptional<z.ZodEnum<{
348
+ anger: "anger";
349
+ happy: "happy";
350
+ neutral: "neutral";
351
+ sad: "sad";
352
+ }>>;
353
+ pace: z.ZodOptional<z.ZodEnum<{
354
+ slow: "slow";
355
+ moderate: "moderate";
356
+ fast: "fast";
357
+ }>>;
358
+ instruct: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
359
+ dialect: z.ZodOptional<z.ZodEnum<{
360
+ cantonese: "cantonese";
361
+ northeastern: "northeastern";
362
+ gansu: "gansu";
363
+ guizhou: "guizhou";
364
+ henan: "henan";
365
+ hubei: "hubei";
366
+ hunan: "hunan";
367
+ jiangxi: "jiangxi";
368
+ minnan: "minnan";
369
+ ningxia: "ningxia";
370
+ shanxi: "shanxi";
371
+ shaanxi: "shaanxi";
372
+ shandong: "shandong";
373
+ shanghai: "shanghai";
374
+ sichuan: "sichuan";
375
+ tianjin: "tianjin";
376
+ yunnan: "yunnan";
377
+ }>>;
378
+ volume: z.ZodOptional<z.ZodEnum<{
379
+ loud: "loud";
380
+ soft: "soft";
381
+ }>>;
382
+ style: z.ZodOptional<z.ZodEnum<{
383
+ peppa: "peppa";
384
+ robot: "robot";
385
+ }>>;
386
+ }, z.core.$strict>]>>;
387
+ useGPU: z.ZodOptional<z.ZodBoolean>;
388
+ outputSampleRate: z.ZodOptional<z.ZodNumber>;
389
+ streamChunkTokens: z.ZodOptional<z.ZodNumber>;
390
+ streamFirstChunkTokens: z.ZodOptional<z.ZodNumber>;
391
+ threads: z.ZodOptional<z.ZodNumber>;
392
+ nGpuLayers: z.ZodOptional<z.ZodNumber>;
393
+ seed: z.ZodOptional<z.ZodNumber>;
394
+ }, z.core.$strip>, z.ZodObject<{
395
+ ttsEngine: z.ZodLiteral<"audio8">;
396
+ referenceText: z.ZodOptional<z.ZodString>;
397
+ greedy: z.ZodOptional<z.ZodBoolean>;
398
+ temperature: z.ZodOptional<z.ZodNumber>;
399
+ topK: z.ZodOptional<z.ZodNumber>;
400
+ topP: z.ZodOptional<z.ZodNumber>;
401
+ maxFrames: z.ZodOptional<z.ZodNumber>;
402
+ useGPU: z.ZodOptional<z.ZodBoolean>;
403
+ outputSampleRate: z.ZodOptional<z.ZodNumber>;
404
+ threads: z.ZodOptional<z.ZodNumber>;
405
+ nGpuLayers: z.ZodOptional<z.ZodNumber>;
406
+ seed: z.ZodOptional<z.ZodNumber>;
239
407
  }, z.core.$strip>], "ttsEngine">;
240
408
  export declare const ttsChatterboxLoadConfigSchema: z.ZodObject<{
241
409
  ttsEngine: z.ZodLiteral<"chatterbox">;
@@ -660,49 +828,18 @@ export declare const ttsParlerLoadConfigSchema: z.ZodObject<{
660
828
  surprise: "surprise";
661
829
  }>>;
662
830
  pitch: z.ZodOptional<z.ZodString>;
663
- pace: z.ZodOptional<z.ZodString>;
831
+ pace: z.ZodOptional<z.ZodEnum<{
832
+ slow: "slow";
833
+ moderate: "moderate";
834
+ fast: "fast";
835
+ }>>;
664
836
  expressivity: z.ZodOptional<z.ZodString>;
665
837
  noise: z.ZodOptional<z.ZodString>;
666
838
  reverb: z.ZodOptional<z.ZodString>;
667
839
  quality: z.ZodOptional<z.ZodString>;
668
840
  ttsEngine: z.ZodLiteral<"parler">;
669
841
  }, z.core.$strip>;
670
- export declare const ttsLoadConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
671
- ttsEngine: z.ZodLiteral<"chatterbox">;
672
- language: z.ZodEnum<{
673
- en: "en";
674
- ar: "ar";
675
- de: "de";
676
- es: "es";
677
- fi: "fi";
678
- fr: "fr";
679
- it: "it";
680
- ja: "ja";
681
- ko: "ko";
682
- nl: "nl";
683
- pl: "pl";
684
- pt: "pt";
685
- ru: "ru";
686
- zh: "zh";
687
- da: "da";
688
- el: "el";
689
- he: "he";
690
- hi: "hi";
691
- ms: "ms";
692
- no: "no";
693
- sv: "sv";
694
- tr: "tr";
695
- sw: "sw";
696
- }>;
697
- voice: z.ZodOptional<z.ZodString>;
698
- useGPU: z.ZodOptional<z.ZodBoolean>;
699
- streamChunkTokens: z.ZodOptional<z.ZodNumber>;
700
- streamFirstChunkTokens: z.ZodOptional<z.ZodNumber>;
701
- cfmSteps: z.ZodOptional<z.ZodNumber>;
702
- cfgRate: z.ZodOptional<z.ZodNumber>;
703
- threads: z.ZodOptional<z.ZodNumber>;
704
- nGpuLayers: z.ZodOptional<z.ZodNumber>;
705
- seed: z.ZodOptional<z.ZodNumber>;
842
+ export declare const ttsCosyvoice3LoadConfigSchema: z.ZodObject<{
706
843
  lavasrEnhancerModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
707
844
  src: z.ZodString;
708
845
  name: z.ZodOptional<z.ZodString>;
@@ -781,7 +918,96 @@ export declare const ttsLoadConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
781
918
  classification: "classification";
782
919
  }>, z.ZodLiteral<"vad">]>>;
783
920
  }, z.core.$strip>]>>;
784
- s3genModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
921
+ ttsEngine: z.ZodLiteral<"cosyvoice3">;
922
+ emotion: z.ZodOptional<z.ZodEnum<{
923
+ anger: "anger";
924
+ happy: "happy";
925
+ neutral: "neutral";
926
+ sad: "sad";
927
+ }>>;
928
+ pace: z.ZodOptional<z.ZodEnum<{
929
+ slow: "slow";
930
+ moderate: "moderate";
931
+ fast: "fast";
932
+ }>>;
933
+ instruct: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
934
+ dialect: z.ZodOptional<z.ZodEnum<{
935
+ cantonese: "cantonese";
936
+ northeastern: "northeastern";
937
+ gansu: "gansu";
938
+ guizhou: "guizhou";
939
+ henan: "henan";
940
+ hubei: "hubei";
941
+ hunan: "hunan";
942
+ jiangxi: "jiangxi";
943
+ minnan: "minnan";
944
+ ningxia: "ningxia";
945
+ shanxi: "shanxi";
946
+ shaanxi: "shaanxi";
947
+ shandong: "shandong";
948
+ shanghai: "shanghai";
949
+ sichuan: "sichuan";
950
+ tianjin: "tianjin";
951
+ yunnan: "yunnan";
952
+ }>>;
953
+ volume: z.ZodOptional<z.ZodEnum<{
954
+ loud: "loud";
955
+ soft: "soft";
956
+ }>>;
957
+ style: z.ZodOptional<z.ZodEnum<{
958
+ peppa: "peppa";
959
+ robot: "robot";
960
+ }>>;
961
+ }, z.core.$strict>]>>;
962
+ useGPU: z.ZodOptional<z.ZodBoolean>;
963
+ outputSampleRate: z.ZodOptional<z.ZodNumber>;
964
+ streamChunkTokens: z.ZodOptional<z.ZodNumber>;
965
+ streamFirstChunkTokens: z.ZodOptional<z.ZodNumber>;
966
+ threads: z.ZodOptional<z.ZodNumber>;
967
+ nGpuLayers: z.ZodOptional<z.ZodNumber>;
968
+ seed: z.ZodOptional<z.ZodNumber>;
969
+ }, z.core.$strip>;
970
+ export declare const ttsAudio8LoadConfigSchema: z.ZodObject<{
971
+ audio8CodecDecoderModelSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
972
+ src: z.ZodString;
973
+ name: z.ZodOptional<z.ZodString>;
974
+ modelId: z.ZodOptional<z.ZodString>;
975
+ registryPath: z.ZodOptional<z.ZodString>;
976
+ registrySource: z.ZodOptional<z.ZodString>;
977
+ blobCoreKey: z.ZodOptional<z.ZodString>;
978
+ blobIndex: z.ZodOptional<z.ZodNumber>;
979
+ engine: z.ZodOptional<z.ZodString>;
980
+ expectedSize: z.ZodOptional<z.ZodNumber>;
981
+ sha256Checksum: z.ZodOptional<z.ZodString>;
982
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
983
+ embeddings: "embeddings";
984
+ diffusion: "diffusion";
985
+ audiogen: "audiogen";
986
+ tts: "tts";
987
+ ocr: "ocr";
988
+ vla: "vla";
989
+ "llamacpp-completion": "llamacpp-completion";
990
+ "whispercpp-transcription": "whispercpp-transcription";
991
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
992
+ "llamacpp-embedding": "llamacpp-embedding";
993
+ "nmtcpp-translation": "nmtcpp-translation";
994
+ "onnx-tts": "onnx-tts";
995
+ "tts-ggml": "tts-ggml";
996
+ "parakeet-transcription": "parakeet-transcription";
997
+ "ggml-ocr": "ggml-ocr";
998
+ "sdcpp-generation": "sdcpp-generation";
999
+ "audiogen-ggml": "audiogen-ggml";
1000
+ "ggml-vla": "ggml-vla";
1001
+ "ggml-classification": "ggml-classification";
1002
+ llm: "llm";
1003
+ whisper: "whisper";
1004
+ bci: "bci";
1005
+ nmt: "nmt";
1006
+ parakeet: "parakeet";
1007
+ classification: "classification";
1008
+ }>, z.ZodLiteral<"vad">]>>;
1009
+ }, z.core.$strip>]>;
1010
+ audio8CodecEncoderModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
785
1011
  src: z.ZodString;
786
1012
  name: z.ZodOptional<z.ZodString>;
787
1013
  modelId: z.ZodOptional<z.ZodString>;
@@ -859,7 +1085,56 @@ export declare const ttsLoadConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
859
1085
  classification: "classification";
860
1086
  }>, z.ZodLiteral<"vad">]>>;
861
1087
  }, z.core.$strip>]>>;
862
- mecabDictSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1088
+ ttsEngine: z.ZodLiteral<"audio8">;
1089
+ referenceText: z.ZodOptional<z.ZodString>;
1090
+ greedy: z.ZodOptional<z.ZodBoolean>;
1091
+ temperature: z.ZodOptional<z.ZodNumber>;
1092
+ topK: z.ZodOptional<z.ZodNumber>;
1093
+ topP: z.ZodOptional<z.ZodNumber>;
1094
+ maxFrames: z.ZodOptional<z.ZodNumber>;
1095
+ useGPU: z.ZodOptional<z.ZodBoolean>;
1096
+ outputSampleRate: z.ZodOptional<z.ZodNumber>;
1097
+ threads: z.ZodOptional<z.ZodNumber>;
1098
+ nGpuLayers: z.ZodOptional<z.ZodNumber>;
1099
+ seed: z.ZodOptional<z.ZodNumber>;
1100
+ }, z.core.$strip>;
1101
+ export declare const ttsLoadConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
1102
+ ttsEngine: z.ZodLiteral<"chatterbox">;
1103
+ language: z.ZodEnum<{
1104
+ en: "en";
1105
+ ar: "ar";
1106
+ de: "de";
1107
+ es: "es";
1108
+ fi: "fi";
1109
+ fr: "fr";
1110
+ it: "it";
1111
+ ja: "ja";
1112
+ ko: "ko";
1113
+ nl: "nl";
1114
+ pl: "pl";
1115
+ pt: "pt";
1116
+ ru: "ru";
1117
+ zh: "zh";
1118
+ da: "da";
1119
+ el: "el";
1120
+ he: "he";
1121
+ hi: "hi";
1122
+ ms: "ms";
1123
+ no: "no";
1124
+ sv: "sv";
1125
+ tr: "tr";
1126
+ sw: "sw";
1127
+ }>;
1128
+ voice: z.ZodOptional<z.ZodString>;
1129
+ useGPU: z.ZodOptional<z.ZodBoolean>;
1130
+ streamChunkTokens: z.ZodOptional<z.ZodNumber>;
1131
+ streamFirstChunkTokens: z.ZodOptional<z.ZodNumber>;
1132
+ cfmSteps: z.ZodOptional<z.ZodNumber>;
1133
+ cfgRate: z.ZodOptional<z.ZodNumber>;
1134
+ threads: z.ZodOptional<z.ZodNumber>;
1135
+ nGpuLayers: z.ZodOptional<z.ZodNumber>;
1136
+ seed: z.ZodOptional<z.ZodNumber>;
1137
+ lavasrEnhancerModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
863
1138
  src: z.ZodString;
864
1139
  name: z.ZodOptional<z.ZodString>;
865
1140
  modelId: z.ZodOptional<z.ZodString>;
@@ -898,7 +1173,7 @@ export declare const ttsLoadConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
898
1173
  classification: "classification";
899
1174
  }>, z.ZodLiteral<"vad">]>>;
900
1175
  }, z.core.$strip>]>>;
901
- cangjieTsvSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1176
+ lavasrDenoiserModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
902
1177
  src: z.ZodString;
903
1178
  name: z.ZodOptional<z.ZodString>;
904
1179
  modelId: z.ZodOptional<z.ZodString>;
@@ -937,48 +1212,7 @@ export declare const ttsLoadConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
937
1212
  classification: "classification";
938
1213
  }>, z.ZodLiteral<"vad">]>>;
939
1214
  }, z.core.$strip>]>>;
940
- }, z.core.$strip>, z.ZodObject<{
941
- ttsEngine: z.ZodLiteral<"supertonic">;
942
- language: z.ZodEnum<{
943
- id: "id";
944
- en: "en";
945
- ar: "ar";
946
- bg: "bg";
947
- cs: "cs";
948
- de: "de";
949
- es: "es";
950
- et: "et";
951
- fi: "fi";
952
- fr: "fr";
953
- hu: "hu";
954
- it: "it";
955
- ja: "ja";
956
- ko: "ko";
957
- lt: "lt";
958
- lv: "lv";
959
- nl: "nl";
960
- pl: "pl";
961
- pt: "pt";
962
- ru: "ru";
963
- sk: "sk";
964
- sl: "sl";
965
- uk: "uk";
966
- da: "da";
967
- el: "el";
968
- hi: "hi";
969
- hr: "hr";
970
- ro: "ro";
971
- sv: "sv";
972
- tr: "tr";
973
- vi: "vi";
974
- }>;
975
- voice: z.ZodOptional<z.ZodString>;
976
- ttsSpeed: z.ZodOptional<z.ZodNumber>;
977
- ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
978
- useGPU: z.ZodOptional<z.ZodBoolean>;
979
- outputSampleRate: z.ZodOptional<z.ZodNumber>;
980
- vulkanCacheDir: z.ZodOptional<z.ZodString>;
981
- lavasrEnhancerModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1215
+ s3genModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
982
1216
  src: z.ZodString;
983
1217
  name: z.ZodOptional<z.ZodString>;
984
1218
  modelId: z.ZodOptional<z.ZodString>;
@@ -1017,7 +1251,7 @@ export declare const ttsLoadConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
1017
1251
  classification: "classification";
1018
1252
  }>, z.ZodLiteral<"vad">]>>;
1019
1253
  }, z.core.$strip>]>>;
1020
- lavasrDenoiserModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1254
+ referenceAudioSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1021
1255
  src: z.ZodString;
1022
1256
  name: z.ZodOptional<z.ZodString>;
1023
1257
  modelId: z.ZodOptional<z.ZodString>;
@@ -1056,25 +1290,222 @@ export declare const ttsLoadConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
1056
1290
  classification: "classification";
1057
1291
  }>, z.ZodLiteral<"vad">]>>;
1058
1292
  }, z.core.$strip>]>>;
1059
- }, z.core.$strip>, z.ZodObject<{
1060
- useGPU: z.ZodOptional<z.ZodBoolean>;
1061
- outputSampleRate: z.ZodOptional<z.ZodNumber>;
1062
- streamChunkTokens: z.ZodOptional<z.ZodNumber>;
1063
- streamFirstChunkTokens: z.ZodOptional<z.ZodNumber>;
1064
- threads: z.ZodOptional<z.ZodNumber>;
1065
- nGpuLayers: z.ZodOptional<z.ZodNumber>;
1066
- seed: z.ZodOptional<z.ZodNumber>;
1067
- temperature: z.ZodOptional<z.ZodNumber>;
1068
- topK: z.ZodOptional<z.ZodNumber>;
1069
- topP: z.ZodOptional<z.ZodNumber>;
1070
- maxFrames: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<0>, z.ZodNumber]>>;
1071
- minNewTokens: z.ZodOptional<z.ZodNumber>;
1072
- normalizeNumbers: z.ZodOptional<z.ZodBoolean>;
1073
- description: z.ZodOptional<z.ZodString>;
1074
- voiceDescription: z.ZodOptional<z.ZodString>;
1075
- voice: z.ZodOptional<z.ZodString>;
1076
- emotion: z.ZodOptional<z.ZodEnum<{
1077
- command: "command";
1293
+ mecabDictSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1294
+ src: z.ZodString;
1295
+ name: z.ZodOptional<z.ZodString>;
1296
+ modelId: z.ZodOptional<z.ZodString>;
1297
+ registryPath: z.ZodOptional<z.ZodString>;
1298
+ registrySource: z.ZodOptional<z.ZodString>;
1299
+ blobCoreKey: z.ZodOptional<z.ZodString>;
1300
+ blobIndex: z.ZodOptional<z.ZodNumber>;
1301
+ engine: z.ZodOptional<z.ZodString>;
1302
+ expectedSize: z.ZodOptional<z.ZodNumber>;
1303
+ sha256Checksum: z.ZodOptional<z.ZodString>;
1304
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
1305
+ embeddings: "embeddings";
1306
+ diffusion: "diffusion";
1307
+ audiogen: "audiogen";
1308
+ tts: "tts";
1309
+ ocr: "ocr";
1310
+ vla: "vla";
1311
+ "llamacpp-completion": "llamacpp-completion";
1312
+ "whispercpp-transcription": "whispercpp-transcription";
1313
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
1314
+ "llamacpp-embedding": "llamacpp-embedding";
1315
+ "nmtcpp-translation": "nmtcpp-translation";
1316
+ "onnx-tts": "onnx-tts";
1317
+ "tts-ggml": "tts-ggml";
1318
+ "parakeet-transcription": "parakeet-transcription";
1319
+ "ggml-ocr": "ggml-ocr";
1320
+ "sdcpp-generation": "sdcpp-generation";
1321
+ "audiogen-ggml": "audiogen-ggml";
1322
+ "ggml-vla": "ggml-vla";
1323
+ "ggml-classification": "ggml-classification";
1324
+ llm: "llm";
1325
+ whisper: "whisper";
1326
+ bci: "bci";
1327
+ nmt: "nmt";
1328
+ parakeet: "parakeet";
1329
+ classification: "classification";
1330
+ }>, z.ZodLiteral<"vad">]>>;
1331
+ }, z.core.$strip>]>>;
1332
+ cangjieTsvSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1333
+ src: z.ZodString;
1334
+ name: z.ZodOptional<z.ZodString>;
1335
+ modelId: z.ZodOptional<z.ZodString>;
1336
+ registryPath: z.ZodOptional<z.ZodString>;
1337
+ registrySource: z.ZodOptional<z.ZodString>;
1338
+ blobCoreKey: z.ZodOptional<z.ZodString>;
1339
+ blobIndex: z.ZodOptional<z.ZodNumber>;
1340
+ engine: z.ZodOptional<z.ZodString>;
1341
+ expectedSize: z.ZodOptional<z.ZodNumber>;
1342
+ sha256Checksum: z.ZodOptional<z.ZodString>;
1343
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
1344
+ embeddings: "embeddings";
1345
+ diffusion: "diffusion";
1346
+ audiogen: "audiogen";
1347
+ tts: "tts";
1348
+ ocr: "ocr";
1349
+ vla: "vla";
1350
+ "llamacpp-completion": "llamacpp-completion";
1351
+ "whispercpp-transcription": "whispercpp-transcription";
1352
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
1353
+ "llamacpp-embedding": "llamacpp-embedding";
1354
+ "nmtcpp-translation": "nmtcpp-translation";
1355
+ "onnx-tts": "onnx-tts";
1356
+ "tts-ggml": "tts-ggml";
1357
+ "parakeet-transcription": "parakeet-transcription";
1358
+ "ggml-ocr": "ggml-ocr";
1359
+ "sdcpp-generation": "sdcpp-generation";
1360
+ "audiogen-ggml": "audiogen-ggml";
1361
+ "ggml-vla": "ggml-vla";
1362
+ "ggml-classification": "ggml-classification";
1363
+ llm: "llm";
1364
+ whisper: "whisper";
1365
+ bci: "bci";
1366
+ nmt: "nmt";
1367
+ parakeet: "parakeet";
1368
+ classification: "classification";
1369
+ }>, z.ZodLiteral<"vad">]>>;
1370
+ }, z.core.$strip>]>>;
1371
+ }, z.core.$strip>, z.ZodObject<{
1372
+ ttsEngine: z.ZodLiteral<"supertonic">;
1373
+ language: z.ZodEnum<{
1374
+ id: "id";
1375
+ en: "en";
1376
+ ar: "ar";
1377
+ bg: "bg";
1378
+ cs: "cs";
1379
+ de: "de";
1380
+ es: "es";
1381
+ et: "et";
1382
+ fi: "fi";
1383
+ fr: "fr";
1384
+ hu: "hu";
1385
+ it: "it";
1386
+ ja: "ja";
1387
+ ko: "ko";
1388
+ lt: "lt";
1389
+ lv: "lv";
1390
+ nl: "nl";
1391
+ pl: "pl";
1392
+ pt: "pt";
1393
+ ru: "ru";
1394
+ sk: "sk";
1395
+ sl: "sl";
1396
+ uk: "uk";
1397
+ da: "da";
1398
+ el: "el";
1399
+ hi: "hi";
1400
+ hr: "hr";
1401
+ ro: "ro";
1402
+ sv: "sv";
1403
+ tr: "tr";
1404
+ vi: "vi";
1405
+ }>;
1406
+ voice: z.ZodOptional<z.ZodString>;
1407
+ ttsSpeed: z.ZodOptional<z.ZodNumber>;
1408
+ ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
1409
+ useGPU: z.ZodOptional<z.ZodBoolean>;
1410
+ outputSampleRate: z.ZodOptional<z.ZodNumber>;
1411
+ vulkanCacheDir: z.ZodOptional<z.ZodString>;
1412
+ lavasrEnhancerModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1413
+ src: z.ZodString;
1414
+ name: z.ZodOptional<z.ZodString>;
1415
+ modelId: z.ZodOptional<z.ZodString>;
1416
+ registryPath: z.ZodOptional<z.ZodString>;
1417
+ registrySource: z.ZodOptional<z.ZodString>;
1418
+ blobCoreKey: z.ZodOptional<z.ZodString>;
1419
+ blobIndex: z.ZodOptional<z.ZodNumber>;
1420
+ engine: z.ZodOptional<z.ZodString>;
1421
+ expectedSize: z.ZodOptional<z.ZodNumber>;
1422
+ sha256Checksum: z.ZodOptional<z.ZodString>;
1423
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
1424
+ embeddings: "embeddings";
1425
+ diffusion: "diffusion";
1426
+ audiogen: "audiogen";
1427
+ tts: "tts";
1428
+ ocr: "ocr";
1429
+ vla: "vla";
1430
+ "llamacpp-completion": "llamacpp-completion";
1431
+ "whispercpp-transcription": "whispercpp-transcription";
1432
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
1433
+ "llamacpp-embedding": "llamacpp-embedding";
1434
+ "nmtcpp-translation": "nmtcpp-translation";
1435
+ "onnx-tts": "onnx-tts";
1436
+ "tts-ggml": "tts-ggml";
1437
+ "parakeet-transcription": "parakeet-transcription";
1438
+ "ggml-ocr": "ggml-ocr";
1439
+ "sdcpp-generation": "sdcpp-generation";
1440
+ "audiogen-ggml": "audiogen-ggml";
1441
+ "ggml-vla": "ggml-vla";
1442
+ "ggml-classification": "ggml-classification";
1443
+ llm: "llm";
1444
+ whisper: "whisper";
1445
+ bci: "bci";
1446
+ nmt: "nmt";
1447
+ parakeet: "parakeet";
1448
+ classification: "classification";
1449
+ }>, z.ZodLiteral<"vad">]>>;
1450
+ }, z.core.$strip>]>>;
1451
+ lavasrDenoiserModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1452
+ src: z.ZodString;
1453
+ name: z.ZodOptional<z.ZodString>;
1454
+ modelId: z.ZodOptional<z.ZodString>;
1455
+ registryPath: z.ZodOptional<z.ZodString>;
1456
+ registrySource: z.ZodOptional<z.ZodString>;
1457
+ blobCoreKey: z.ZodOptional<z.ZodString>;
1458
+ blobIndex: z.ZodOptional<z.ZodNumber>;
1459
+ engine: z.ZodOptional<z.ZodString>;
1460
+ expectedSize: z.ZodOptional<z.ZodNumber>;
1461
+ sha256Checksum: z.ZodOptional<z.ZodString>;
1462
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
1463
+ embeddings: "embeddings";
1464
+ diffusion: "diffusion";
1465
+ audiogen: "audiogen";
1466
+ tts: "tts";
1467
+ ocr: "ocr";
1468
+ vla: "vla";
1469
+ "llamacpp-completion": "llamacpp-completion";
1470
+ "whispercpp-transcription": "whispercpp-transcription";
1471
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
1472
+ "llamacpp-embedding": "llamacpp-embedding";
1473
+ "nmtcpp-translation": "nmtcpp-translation";
1474
+ "onnx-tts": "onnx-tts";
1475
+ "tts-ggml": "tts-ggml";
1476
+ "parakeet-transcription": "parakeet-transcription";
1477
+ "ggml-ocr": "ggml-ocr";
1478
+ "sdcpp-generation": "sdcpp-generation";
1479
+ "audiogen-ggml": "audiogen-ggml";
1480
+ "ggml-vla": "ggml-vla";
1481
+ "ggml-classification": "ggml-classification";
1482
+ llm: "llm";
1483
+ whisper: "whisper";
1484
+ bci: "bci";
1485
+ nmt: "nmt";
1486
+ parakeet: "parakeet";
1487
+ classification: "classification";
1488
+ }>, z.ZodLiteral<"vad">]>>;
1489
+ }, z.core.$strip>]>>;
1490
+ }, z.core.$strip>, z.ZodObject<{
1491
+ useGPU: z.ZodOptional<z.ZodBoolean>;
1492
+ outputSampleRate: z.ZodOptional<z.ZodNumber>;
1493
+ streamChunkTokens: z.ZodOptional<z.ZodNumber>;
1494
+ streamFirstChunkTokens: z.ZodOptional<z.ZodNumber>;
1495
+ threads: z.ZodOptional<z.ZodNumber>;
1496
+ nGpuLayers: z.ZodOptional<z.ZodNumber>;
1497
+ seed: z.ZodOptional<z.ZodNumber>;
1498
+ temperature: z.ZodOptional<z.ZodNumber>;
1499
+ topK: z.ZodOptional<z.ZodNumber>;
1500
+ topP: z.ZodOptional<z.ZodNumber>;
1501
+ maxFrames: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<0>, z.ZodNumber]>>;
1502
+ minNewTokens: z.ZodOptional<z.ZodNumber>;
1503
+ normalizeNumbers: z.ZodOptional<z.ZodBoolean>;
1504
+ description: z.ZodOptional<z.ZodString>;
1505
+ voiceDescription: z.ZodOptional<z.ZodString>;
1506
+ voice: z.ZodOptional<z.ZodString>;
1507
+ emotion: z.ZodOptional<z.ZodEnum<{
1508
+ command: "command";
1078
1509
  anger: "anger";
1079
1510
  narration: "narration";
1080
1511
  conversation: "conversation";
@@ -1088,12 +1519,273 @@ export declare const ttsLoadConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
1088
1519
  surprise: "surprise";
1089
1520
  }>>;
1090
1521
  pitch: z.ZodOptional<z.ZodString>;
1091
- pace: z.ZodOptional<z.ZodString>;
1522
+ pace: z.ZodOptional<z.ZodEnum<{
1523
+ slow: "slow";
1524
+ moderate: "moderate";
1525
+ fast: "fast";
1526
+ }>>;
1092
1527
  expressivity: z.ZodOptional<z.ZodString>;
1093
1528
  noise: z.ZodOptional<z.ZodString>;
1094
1529
  reverb: z.ZodOptional<z.ZodString>;
1095
1530
  quality: z.ZodOptional<z.ZodString>;
1096
1531
  ttsEngine: z.ZodLiteral<"parler">;
1532
+ }, z.core.$strip>, z.ZodObject<{
1533
+ lavasrEnhancerModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1534
+ src: z.ZodString;
1535
+ name: z.ZodOptional<z.ZodString>;
1536
+ modelId: z.ZodOptional<z.ZodString>;
1537
+ registryPath: z.ZodOptional<z.ZodString>;
1538
+ registrySource: z.ZodOptional<z.ZodString>;
1539
+ blobCoreKey: z.ZodOptional<z.ZodString>;
1540
+ blobIndex: z.ZodOptional<z.ZodNumber>;
1541
+ engine: z.ZodOptional<z.ZodString>;
1542
+ expectedSize: z.ZodOptional<z.ZodNumber>;
1543
+ sha256Checksum: z.ZodOptional<z.ZodString>;
1544
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
1545
+ embeddings: "embeddings";
1546
+ diffusion: "diffusion";
1547
+ audiogen: "audiogen";
1548
+ tts: "tts";
1549
+ ocr: "ocr";
1550
+ vla: "vla";
1551
+ "llamacpp-completion": "llamacpp-completion";
1552
+ "whispercpp-transcription": "whispercpp-transcription";
1553
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
1554
+ "llamacpp-embedding": "llamacpp-embedding";
1555
+ "nmtcpp-translation": "nmtcpp-translation";
1556
+ "onnx-tts": "onnx-tts";
1557
+ "tts-ggml": "tts-ggml";
1558
+ "parakeet-transcription": "parakeet-transcription";
1559
+ "ggml-ocr": "ggml-ocr";
1560
+ "sdcpp-generation": "sdcpp-generation";
1561
+ "audiogen-ggml": "audiogen-ggml";
1562
+ "ggml-vla": "ggml-vla";
1563
+ "ggml-classification": "ggml-classification";
1564
+ llm: "llm";
1565
+ whisper: "whisper";
1566
+ bci: "bci";
1567
+ nmt: "nmt";
1568
+ parakeet: "parakeet";
1569
+ classification: "classification";
1570
+ }>, z.ZodLiteral<"vad">]>>;
1571
+ }, z.core.$strip>]>>;
1572
+ lavasrDenoiserModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1573
+ src: z.ZodString;
1574
+ name: z.ZodOptional<z.ZodString>;
1575
+ modelId: z.ZodOptional<z.ZodString>;
1576
+ registryPath: z.ZodOptional<z.ZodString>;
1577
+ registrySource: z.ZodOptional<z.ZodString>;
1578
+ blobCoreKey: z.ZodOptional<z.ZodString>;
1579
+ blobIndex: z.ZodOptional<z.ZodNumber>;
1580
+ engine: z.ZodOptional<z.ZodString>;
1581
+ expectedSize: z.ZodOptional<z.ZodNumber>;
1582
+ sha256Checksum: z.ZodOptional<z.ZodString>;
1583
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
1584
+ embeddings: "embeddings";
1585
+ diffusion: "diffusion";
1586
+ audiogen: "audiogen";
1587
+ tts: "tts";
1588
+ ocr: "ocr";
1589
+ vla: "vla";
1590
+ "llamacpp-completion": "llamacpp-completion";
1591
+ "whispercpp-transcription": "whispercpp-transcription";
1592
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
1593
+ "llamacpp-embedding": "llamacpp-embedding";
1594
+ "nmtcpp-translation": "nmtcpp-translation";
1595
+ "onnx-tts": "onnx-tts";
1596
+ "tts-ggml": "tts-ggml";
1597
+ "parakeet-transcription": "parakeet-transcription";
1598
+ "ggml-ocr": "ggml-ocr";
1599
+ "sdcpp-generation": "sdcpp-generation";
1600
+ "audiogen-ggml": "audiogen-ggml";
1601
+ "ggml-vla": "ggml-vla";
1602
+ "ggml-classification": "ggml-classification";
1603
+ llm: "llm";
1604
+ whisper: "whisper";
1605
+ bci: "bci";
1606
+ nmt: "nmt";
1607
+ parakeet: "parakeet";
1608
+ classification: "classification";
1609
+ }>, z.ZodLiteral<"vad">]>>;
1610
+ }, z.core.$strip>]>>;
1611
+ ttsEngine: z.ZodLiteral<"cosyvoice3">;
1612
+ emotion: z.ZodOptional<z.ZodEnum<{
1613
+ anger: "anger";
1614
+ happy: "happy";
1615
+ neutral: "neutral";
1616
+ sad: "sad";
1617
+ }>>;
1618
+ pace: z.ZodOptional<z.ZodEnum<{
1619
+ slow: "slow";
1620
+ moderate: "moderate";
1621
+ fast: "fast";
1622
+ }>>;
1623
+ instruct: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1624
+ dialect: z.ZodOptional<z.ZodEnum<{
1625
+ cantonese: "cantonese";
1626
+ northeastern: "northeastern";
1627
+ gansu: "gansu";
1628
+ guizhou: "guizhou";
1629
+ henan: "henan";
1630
+ hubei: "hubei";
1631
+ hunan: "hunan";
1632
+ jiangxi: "jiangxi";
1633
+ minnan: "minnan";
1634
+ ningxia: "ningxia";
1635
+ shanxi: "shanxi";
1636
+ shaanxi: "shaanxi";
1637
+ shandong: "shandong";
1638
+ shanghai: "shanghai";
1639
+ sichuan: "sichuan";
1640
+ tianjin: "tianjin";
1641
+ yunnan: "yunnan";
1642
+ }>>;
1643
+ volume: z.ZodOptional<z.ZodEnum<{
1644
+ loud: "loud";
1645
+ soft: "soft";
1646
+ }>>;
1647
+ style: z.ZodOptional<z.ZodEnum<{
1648
+ peppa: "peppa";
1649
+ robot: "robot";
1650
+ }>>;
1651
+ }, z.core.$strict>]>>;
1652
+ useGPU: z.ZodOptional<z.ZodBoolean>;
1653
+ outputSampleRate: z.ZodOptional<z.ZodNumber>;
1654
+ streamChunkTokens: z.ZodOptional<z.ZodNumber>;
1655
+ streamFirstChunkTokens: z.ZodOptional<z.ZodNumber>;
1656
+ threads: z.ZodOptional<z.ZodNumber>;
1657
+ nGpuLayers: z.ZodOptional<z.ZodNumber>;
1658
+ seed: z.ZodOptional<z.ZodNumber>;
1659
+ }, z.core.$strip>, z.ZodObject<{
1660
+ audio8CodecDecoderModelSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1661
+ src: z.ZodString;
1662
+ name: z.ZodOptional<z.ZodString>;
1663
+ modelId: z.ZodOptional<z.ZodString>;
1664
+ registryPath: z.ZodOptional<z.ZodString>;
1665
+ registrySource: z.ZodOptional<z.ZodString>;
1666
+ blobCoreKey: z.ZodOptional<z.ZodString>;
1667
+ blobIndex: z.ZodOptional<z.ZodNumber>;
1668
+ engine: z.ZodOptional<z.ZodString>;
1669
+ expectedSize: z.ZodOptional<z.ZodNumber>;
1670
+ sha256Checksum: z.ZodOptional<z.ZodString>;
1671
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
1672
+ embeddings: "embeddings";
1673
+ diffusion: "diffusion";
1674
+ audiogen: "audiogen";
1675
+ tts: "tts";
1676
+ ocr: "ocr";
1677
+ vla: "vla";
1678
+ "llamacpp-completion": "llamacpp-completion";
1679
+ "whispercpp-transcription": "whispercpp-transcription";
1680
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
1681
+ "llamacpp-embedding": "llamacpp-embedding";
1682
+ "nmtcpp-translation": "nmtcpp-translation";
1683
+ "onnx-tts": "onnx-tts";
1684
+ "tts-ggml": "tts-ggml";
1685
+ "parakeet-transcription": "parakeet-transcription";
1686
+ "ggml-ocr": "ggml-ocr";
1687
+ "sdcpp-generation": "sdcpp-generation";
1688
+ "audiogen-ggml": "audiogen-ggml";
1689
+ "ggml-vla": "ggml-vla";
1690
+ "ggml-classification": "ggml-classification";
1691
+ llm: "llm";
1692
+ whisper: "whisper";
1693
+ bci: "bci";
1694
+ nmt: "nmt";
1695
+ parakeet: "parakeet";
1696
+ classification: "classification";
1697
+ }>, z.ZodLiteral<"vad">]>>;
1698
+ }, z.core.$strip>]>;
1699
+ audio8CodecEncoderModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1700
+ src: z.ZodString;
1701
+ name: z.ZodOptional<z.ZodString>;
1702
+ modelId: z.ZodOptional<z.ZodString>;
1703
+ registryPath: z.ZodOptional<z.ZodString>;
1704
+ registrySource: z.ZodOptional<z.ZodString>;
1705
+ blobCoreKey: z.ZodOptional<z.ZodString>;
1706
+ blobIndex: z.ZodOptional<z.ZodNumber>;
1707
+ engine: z.ZodOptional<z.ZodString>;
1708
+ expectedSize: z.ZodOptional<z.ZodNumber>;
1709
+ sha256Checksum: z.ZodOptional<z.ZodString>;
1710
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
1711
+ embeddings: "embeddings";
1712
+ diffusion: "diffusion";
1713
+ audiogen: "audiogen";
1714
+ tts: "tts";
1715
+ ocr: "ocr";
1716
+ vla: "vla";
1717
+ "llamacpp-completion": "llamacpp-completion";
1718
+ "whispercpp-transcription": "whispercpp-transcription";
1719
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
1720
+ "llamacpp-embedding": "llamacpp-embedding";
1721
+ "nmtcpp-translation": "nmtcpp-translation";
1722
+ "onnx-tts": "onnx-tts";
1723
+ "tts-ggml": "tts-ggml";
1724
+ "parakeet-transcription": "parakeet-transcription";
1725
+ "ggml-ocr": "ggml-ocr";
1726
+ "sdcpp-generation": "sdcpp-generation";
1727
+ "audiogen-ggml": "audiogen-ggml";
1728
+ "ggml-vla": "ggml-vla";
1729
+ "ggml-classification": "ggml-classification";
1730
+ llm: "llm";
1731
+ whisper: "whisper";
1732
+ bci: "bci";
1733
+ nmt: "nmt";
1734
+ parakeet: "parakeet";
1735
+ classification: "classification";
1736
+ }>, z.ZodLiteral<"vad">]>>;
1737
+ }, z.core.$strip>]>>;
1738
+ referenceAudioSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1739
+ src: z.ZodString;
1740
+ name: z.ZodOptional<z.ZodString>;
1741
+ modelId: z.ZodOptional<z.ZodString>;
1742
+ registryPath: z.ZodOptional<z.ZodString>;
1743
+ registrySource: z.ZodOptional<z.ZodString>;
1744
+ blobCoreKey: z.ZodOptional<z.ZodString>;
1745
+ blobIndex: z.ZodOptional<z.ZodNumber>;
1746
+ engine: z.ZodOptional<z.ZodString>;
1747
+ expectedSize: z.ZodOptional<z.ZodNumber>;
1748
+ sha256Checksum: z.ZodOptional<z.ZodString>;
1749
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
1750
+ embeddings: "embeddings";
1751
+ diffusion: "diffusion";
1752
+ audiogen: "audiogen";
1753
+ tts: "tts";
1754
+ ocr: "ocr";
1755
+ vla: "vla";
1756
+ "llamacpp-completion": "llamacpp-completion";
1757
+ "whispercpp-transcription": "whispercpp-transcription";
1758
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
1759
+ "llamacpp-embedding": "llamacpp-embedding";
1760
+ "nmtcpp-translation": "nmtcpp-translation";
1761
+ "onnx-tts": "onnx-tts";
1762
+ "tts-ggml": "tts-ggml";
1763
+ "parakeet-transcription": "parakeet-transcription";
1764
+ "ggml-ocr": "ggml-ocr";
1765
+ "sdcpp-generation": "sdcpp-generation";
1766
+ "audiogen-ggml": "audiogen-ggml";
1767
+ "ggml-vla": "ggml-vla";
1768
+ "ggml-classification": "ggml-classification";
1769
+ llm: "llm";
1770
+ whisper: "whisper";
1771
+ bci: "bci";
1772
+ nmt: "nmt";
1773
+ parakeet: "parakeet";
1774
+ classification: "classification";
1775
+ }>, z.ZodLiteral<"vad">]>>;
1776
+ }, z.core.$strip>]>>;
1777
+ ttsEngine: z.ZodLiteral<"audio8">;
1778
+ referenceText: z.ZodOptional<z.ZodString>;
1779
+ greedy: z.ZodOptional<z.ZodBoolean>;
1780
+ temperature: z.ZodOptional<z.ZodNumber>;
1781
+ topK: z.ZodOptional<z.ZodNumber>;
1782
+ topP: z.ZodOptional<z.ZodNumber>;
1783
+ maxFrames: z.ZodOptional<z.ZodNumber>;
1784
+ useGPU: z.ZodOptional<z.ZodBoolean>;
1785
+ outputSampleRate: z.ZodOptional<z.ZodNumber>;
1786
+ threads: z.ZodOptional<z.ZodNumber>;
1787
+ nGpuLayers: z.ZodOptional<z.ZodNumber>;
1788
+ seed: z.ZodOptional<z.ZodNumber>;
1097
1789
  }, z.core.$strip>], "ttsEngine">;
1098
1790
  export declare const LEGACY_TTS_ONNX_MODEL_CONFIG_FIELDS: readonly ["ttsSupertonicMultilingual", "ttsTokenizerSrc", "ttsSpeechEncoderSrc", "ttsEmbedTokensSrc", "ttsConditionalDecoderSrc", "ttsLanguageModelSrc", "ttsTextEncoderSrc", "ttsDurationPredictorSrc", "ttsVectorEstimatorSrc", "ttsVocoderSrc", "ttsUnicodeIndexerSrc", "ttsTtsConfigSrc", "ttsVoiceStyleSrc"];
1099
1791
  export declare const ttsConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
@@ -1132,12 +1824,273 @@ export declare const ttsConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
1132
1824
  surprise: "surprise";
1133
1825
  }>>;
1134
1826
  pitch: z.ZodOptional<z.ZodString>;
1135
- pace: z.ZodOptional<z.ZodString>;
1827
+ pace: z.ZodOptional<z.ZodEnum<{
1828
+ slow: "slow";
1829
+ moderate: "moderate";
1830
+ fast: "fast";
1831
+ }>>;
1136
1832
  expressivity: z.ZodOptional<z.ZodString>;
1137
1833
  noise: z.ZodOptional<z.ZodString>;
1138
1834
  reverb: z.ZodOptional<z.ZodString>;
1139
1835
  quality: z.ZodOptional<z.ZodString>;
1140
1836
  ttsEngine: z.ZodLiteral<"parler">;
1837
+ }, z.core.$strict>, z.ZodObject<{
1838
+ lavasrEnhancerModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1839
+ src: z.ZodString;
1840
+ name: z.ZodOptional<z.ZodString>;
1841
+ modelId: z.ZodOptional<z.ZodString>;
1842
+ registryPath: z.ZodOptional<z.ZodString>;
1843
+ registrySource: z.ZodOptional<z.ZodString>;
1844
+ blobCoreKey: z.ZodOptional<z.ZodString>;
1845
+ blobIndex: z.ZodOptional<z.ZodNumber>;
1846
+ engine: z.ZodOptional<z.ZodString>;
1847
+ expectedSize: z.ZodOptional<z.ZodNumber>;
1848
+ sha256Checksum: z.ZodOptional<z.ZodString>;
1849
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
1850
+ embeddings: "embeddings";
1851
+ diffusion: "diffusion";
1852
+ audiogen: "audiogen";
1853
+ tts: "tts";
1854
+ ocr: "ocr";
1855
+ vla: "vla";
1856
+ "llamacpp-completion": "llamacpp-completion";
1857
+ "whispercpp-transcription": "whispercpp-transcription";
1858
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
1859
+ "llamacpp-embedding": "llamacpp-embedding";
1860
+ "nmtcpp-translation": "nmtcpp-translation";
1861
+ "onnx-tts": "onnx-tts";
1862
+ "tts-ggml": "tts-ggml";
1863
+ "parakeet-transcription": "parakeet-transcription";
1864
+ "ggml-ocr": "ggml-ocr";
1865
+ "sdcpp-generation": "sdcpp-generation";
1866
+ "audiogen-ggml": "audiogen-ggml";
1867
+ "ggml-vla": "ggml-vla";
1868
+ "ggml-classification": "ggml-classification";
1869
+ llm: "llm";
1870
+ whisper: "whisper";
1871
+ bci: "bci";
1872
+ nmt: "nmt";
1873
+ parakeet: "parakeet";
1874
+ classification: "classification";
1875
+ }>, z.ZodLiteral<"vad">]>>;
1876
+ }, z.core.$strip>]>>;
1877
+ lavasrDenoiserModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1878
+ src: z.ZodString;
1879
+ name: z.ZodOptional<z.ZodString>;
1880
+ modelId: z.ZodOptional<z.ZodString>;
1881
+ registryPath: z.ZodOptional<z.ZodString>;
1882
+ registrySource: z.ZodOptional<z.ZodString>;
1883
+ blobCoreKey: z.ZodOptional<z.ZodString>;
1884
+ blobIndex: z.ZodOptional<z.ZodNumber>;
1885
+ engine: z.ZodOptional<z.ZodString>;
1886
+ expectedSize: z.ZodOptional<z.ZodNumber>;
1887
+ sha256Checksum: z.ZodOptional<z.ZodString>;
1888
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
1889
+ embeddings: "embeddings";
1890
+ diffusion: "diffusion";
1891
+ audiogen: "audiogen";
1892
+ tts: "tts";
1893
+ ocr: "ocr";
1894
+ vla: "vla";
1895
+ "llamacpp-completion": "llamacpp-completion";
1896
+ "whispercpp-transcription": "whispercpp-transcription";
1897
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
1898
+ "llamacpp-embedding": "llamacpp-embedding";
1899
+ "nmtcpp-translation": "nmtcpp-translation";
1900
+ "onnx-tts": "onnx-tts";
1901
+ "tts-ggml": "tts-ggml";
1902
+ "parakeet-transcription": "parakeet-transcription";
1903
+ "ggml-ocr": "ggml-ocr";
1904
+ "sdcpp-generation": "sdcpp-generation";
1905
+ "audiogen-ggml": "audiogen-ggml";
1906
+ "ggml-vla": "ggml-vla";
1907
+ "ggml-classification": "ggml-classification";
1908
+ llm: "llm";
1909
+ whisper: "whisper";
1910
+ bci: "bci";
1911
+ nmt: "nmt";
1912
+ parakeet: "parakeet";
1913
+ classification: "classification";
1914
+ }>, z.ZodLiteral<"vad">]>>;
1915
+ }, z.core.$strip>]>>;
1916
+ ttsEngine: z.ZodLiteral<"cosyvoice3">;
1917
+ emotion: z.ZodOptional<z.ZodEnum<{
1918
+ anger: "anger";
1919
+ happy: "happy";
1920
+ neutral: "neutral";
1921
+ sad: "sad";
1922
+ }>>;
1923
+ pace: z.ZodOptional<z.ZodEnum<{
1924
+ slow: "slow";
1925
+ moderate: "moderate";
1926
+ fast: "fast";
1927
+ }>>;
1928
+ instruct: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1929
+ dialect: z.ZodOptional<z.ZodEnum<{
1930
+ cantonese: "cantonese";
1931
+ northeastern: "northeastern";
1932
+ gansu: "gansu";
1933
+ guizhou: "guizhou";
1934
+ henan: "henan";
1935
+ hubei: "hubei";
1936
+ hunan: "hunan";
1937
+ jiangxi: "jiangxi";
1938
+ minnan: "minnan";
1939
+ ningxia: "ningxia";
1940
+ shanxi: "shanxi";
1941
+ shaanxi: "shaanxi";
1942
+ shandong: "shandong";
1943
+ shanghai: "shanghai";
1944
+ sichuan: "sichuan";
1945
+ tianjin: "tianjin";
1946
+ yunnan: "yunnan";
1947
+ }>>;
1948
+ volume: z.ZodOptional<z.ZodEnum<{
1949
+ loud: "loud";
1950
+ soft: "soft";
1951
+ }>>;
1952
+ style: z.ZodOptional<z.ZodEnum<{
1953
+ peppa: "peppa";
1954
+ robot: "robot";
1955
+ }>>;
1956
+ }, z.core.$strict>]>>;
1957
+ useGPU: z.ZodOptional<z.ZodBoolean>;
1958
+ outputSampleRate: z.ZodOptional<z.ZodNumber>;
1959
+ streamChunkTokens: z.ZodOptional<z.ZodNumber>;
1960
+ streamFirstChunkTokens: z.ZodOptional<z.ZodNumber>;
1961
+ threads: z.ZodOptional<z.ZodNumber>;
1962
+ nGpuLayers: z.ZodOptional<z.ZodNumber>;
1963
+ seed: z.ZodOptional<z.ZodNumber>;
1964
+ }, z.core.$strict>, z.ZodObject<{
1965
+ audio8CodecDecoderModelSrc: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
1966
+ src: z.ZodString;
1967
+ name: z.ZodOptional<z.ZodString>;
1968
+ modelId: z.ZodOptional<z.ZodString>;
1969
+ registryPath: z.ZodOptional<z.ZodString>;
1970
+ registrySource: z.ZodOptional<z.ZodString>;
1971
+ blobCoreKey: z.ZodOptional<z.ZodString>;
1972
+ blobIndex: z.ZodOptional<z.ZodNumber>;
1973
+ engine: z.ZodOptional<z.ZodString>;
1974
+ expectedSize: z.ZodOptional<z.ZodNumber>;
1975
+ sha256Checksum: z.ZodOptional<z.ZodString>;
1976
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
1977
+ embeddings: "embeddings";
1978
+ diffusion: "diffusion";
1979
+ audiogen: "audiogen";
1980
+ tts: "tts";
1981
+ ocr: "ocr";
1982
+ vla: "vla";
1983
+ "llamacpp-completion": "llamacpp-completion";
1984
+ "whispercpp-transcription": "whispercpp-transcription";
1985
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
1986
+ "llamacpp-embedding": "llamacpp-embedding";
1987
+ "nmtcpp-translation": "nmtcpp-translation";
1988
+ "onnx-tts": "onnx-tts";
1989
+ "tts-ggml": "tts-ggml";
1990
+ "parakeet-transcription": "parakeet-transcription";
1991
+ "ggml-ocr": "ggml-ocr";
1992
+ "sdcpp-generation": "sdcpp-generation";
1993
+ "audiogen-ggml": "audiogen-ggml";
1994
+ "ggml-vla": "ggml-vla";
1995
+ "ggml-classification": "ggml-classification";
1996
+ llm: "llm";
1997
+ whisper: "whisper";
1998
+ bci: "bci";
1999
+ nmt: "nmt";
2000
+ parakeet: "parakeet";
2001
+ classification: "classification";
2002
+ }>, z.ZodLiteral<"vad">]>>;
2003
+ }, z.core.$strip>]>;
2004
+ audio8CodecEncoderModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
2005
+ src: z.ZodString;
2006
+ name: z.ZodOptional<z.ZodString>;
2007
+ modelId: z.ZodOptional<z.ZodString>;
2008
+ registryPath: z.ZodOptional<z.ZodString>;
2009
+ registrySource: z.ZodOptional<z.ZodString>;
2010
+ blobCoreKey: z.ZodOptional<z.ZodString>;
2011
+ blobIndex: z.ZodOptional<z.ZodNumber>;
2012
+ engine: z.ZodOptional<z.ZodString>;
2013
+ expectedSize: z.ZodOptional<z.ZodNumber>;
2014
+ sha256Checksum: z.ZodOptional<z.ZodString>;
2015
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
2016
+ embeddings: "embeddings";
2017
+ diffusion: "diffusion";
2018
+ audiogen: "audiogen";
2019
+ tts: "tts";
2020
+ ocr: "ocr";
2021
+ vla: "vla";
2022
+ "llamacpp-completion": "llamacpp-completion";
2023
+ "whispercpp-transcription": "whispercpp-transcription";
2024
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
2025
+ "llamacpp-embedding": "llamacpp-embedding";
2026
+ "nmtcpp-translation": "nmtcpp-translation";
2027
+ "onnx-tts": "onnx-tts";
2028
+ "tts-ggml": "tts-ggml";
2029
+ "parakeet-transcription": "parakeet-transcription";
2030
+ "ggml-ocr": "ggml-ocr";
2031
+ "sdcpp-generation": "sdcpp-generation";
2032
+ "audiogen-ggml": "audiogen-ggml";
2033
+ "ggml-vla": "ggml-vla";
2034
+ "ggml-classification": "ggml-classification";
2035
+ llm: "llm";
2036
+ whisper: "whisper";
2037
+ bci: "bci";
2038
+ nmt: "nmt";
2039
+ parakeet: "parakeet";
2040
+ classification: "classification";
2041
+ }>, z.ZodLiteral<"vad">]>>;
2042
+ }, z.core.$strip>]>>;
2043
+ referenceAudioSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
2044
+ src: z.ZodString;
2045
+ name: z.ZodOptional<z.ZodString>;
2046
+ modelId: z.ZodOptional<z.ZodString>;
2047
+ registryPath: z.ZodOptional<z.ZodString>;
2048
+ registrySource: z.ZodOptional<z.ZodString>;
2049
+ blobCoreKey: z.ZodOptional<z.ZodString>;
2050
+ blobIndex: z.ZodOptional<z.ZodNumber>;
2051
+ engine: z.ZodOptional<z.ZodString>;
2052
+ expectedSize: z.ZodOptional<z.ZodNumber>;
2053
+ sha256Checksum: z.ZodOptional<z.ZodString>;
2054
+ addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
2055
+ embeddings: "embeddings";
2056
+ diffusion: "diffusion";
2057
+ audiogen: "audiogen";
2058
+ tts: "tts";
2059
+ ocr: "ocr";
2060
+ vla: "vla";
2061
+ "llamacpp-completion": "llamacpp-completion";
2062
+ "whispercpp-transcription": "whispercpp-transcription";
2063
+ "bci-whispercpp-transcription": "bci-whispercpp-transcription";
2064
+ "llamacpp-embedding": "llamacpp-embedding";
2065
+ "nmtcpp-translation": "nmtcpp-translation";
2066
+ "onnx-tts": "onnx-tts";
2067
+ "tts-ggml": "tts-ggml";
2068
+ "parakeet-transcription": "parakeet-transcription";
2069
+ "ggml-ocr": "ggml-ocr";
2070
+ "sdcpp-generation": "sdcpp-generation";
2071
+ "audiogen-ggml": "audiogen-ggml";
2072
+ "ggml-vla": "ggml-vla";
2073
+ "ggml-classification": "ggml-classification";
2074
+ llm: "llm";
2075
+ whisper: "whisper";
2076
+ bci: "bci";
2077
+ nmt: "nmt";
2078
+ parakeet: "parakeet";
2079
+ classification: "classification";
2080
+ }>, z.ZodLiteral<"vad">]>>;
2081
+ }, z.core.$strip>]>>;
2082
+ ttsEngine: z.ZodLiteral<"audio8">;
2083
+ referenceText: z.ZodOptional<z.ZodString>;
2084
+ greedy: z.ZodOptional<z.ZodBoolean>;
2085
+ temperature: z.ZodOptional<z.ZodNumber>;
2086
+ topK: z.ZodOptional<z.ZodNumber>;
2087
+ topP: z.ZodOptional<z.ZodNumber>;
2088
+ maxFrames: z.ZodOptional<z.ZodNumber>;
2089
+ useGPU: z.ZodOptional<z.ZodBoolean>;
2090
+ outputSampleRate: z.ZodOptional<z.ZodNumber>;
2091
+ threads: z.ZodOptional<z.ZodNumber>;
2092
+ nGpuLayers: z.ZodOptional<z.ZodNumber>;
2093
+ seed: z.ZodOptional<z.ZodNumber>;
1141
2094
  }, z.core.$strict>], "ttsEngine">;
1142
2095
  export declare const ttsClientParamsSchema: z.ZodObject<{
1143
2096
  description: z.ZodOptional<z.ZodString>;
@@ -1158,7 +2111,11 @@ export declare const ttsClientParamsSchema: z.ZodObject<{
1158
2111
  surprise: "surprise";
1159
2112
  }>>;
1160
2113
  pitch: z.ZodOptional<z.ZodString>;
1161
- pace: z.ZodOptional<z.ZodString>;
2114
+ pace: z.ZodOptional<z.ZodEnum<{
2115
+ slow: "slow";
2116
+ moderate: "moderate";
2117
+ fast: "fast";
2118
+ }>>;
1162
2119
  expressivity: z.ZodOptional<z.ZodString>;
1163
2120
  noise: z.ZodOptional<z.ZodString>;
1164
2121
  reverb: z.ZodOptional<z.ZodString>;
@@ -1191,7 +2148,11 @@ export declare const ttsRequestSchema: z.ZodObject<{
1191
2148
  surprise: "surprise";
1192
2149
  }>>;
1193
2150
  pitch: z.ZodOptional<z.ZodString>;
1194
- pace: z.ZodOptional<z.ZodString>;
2151
+ pace: z.ZodOptional<z.ZodEnum<{
2152
+ slow: "slow";
2153
+ moderate: "moderate";
2154
+ fast: "fast";
2155
+ }>>;
1195
2156
  expressivity: z.ZodOptional<z.ZodString>;
1196
2157
  noise: z.ZodOptional<z.ZodString>;
1197
2158
  reverb: z.ZodOptional<z.ZodString>;
@@ -1242,7 +2203,11 @@ declare const textToSpeechStreamRequestBaseShape: {
1242
2203
  surprise: "surprise";
1243
2204
  }>>;
1244
2205
  pitch: z.ZodOptional<z.ZodString>;
1245
- pace: z.ZodOptional<z.ZodString>;
2206
+ pace: z.ZodOptional<z.ZodEnum<{
2207
+ slow: "slow";
2208
+ moderate: "moderate";
2209
+ fast: "fast";
2210
+ }>>;
1246
2211
  expressivity: z.ZodOptional<z.ZodString>;
1247
2212
  noise: z.ZodOptional<z.ZodString>;
1248
2213
  reverb: z.ZodOptional<z.ZodString>;
@@ -1278,7 +2243,11 @@ export declare const textToSpeechStreamRequestSchema: z.ZodObject<{
1278
2243
  surprise: "surprise";
1279
2244
  }>>;
1280
2245
  pitch: z.ZodOptional<z.ZodString>;
1281
- pace: z.ZodOptional<z.ZodString>;
2246
+ pace: z.ZodOptional<z.ZodEnum<{
2247
+ slow: "slow";
2248
+ moderate: "moderate";
2249
+ fast: "fast";
2250
+ }>>;
1282
2251
  expressivity: z.ZodOptional<z.ZodString>;
1283
2252
  noise: z.ZodOptional<z.ZodString>;
1284
2253
  reverb: z.ZodOptional<z.ZodString>;
@@ -1311,9 +2280,14 @@ export type TtsLanguage = (typeof TTS_LANGUAGES)[number];
1311
2280
  export type TtsChatterboxLanguage = (typeof TTS_CHATTERBOX_LANGUAGES)[number];
1312
2281
  export type TtsSupertonicLanguage = (typeof TTS_SUPERTONIC_LANGUAGES)[number];
1313
2282
  export type TtsParlerEmotion = (typeof TTS_PARLER_EMOTIONS)[number];
2283
+ export type TtsPace = (typeof TTS_PACES)[number];
2284
+ export type TtsCosyvoice3Emotion = (typeof TTS_COSYVOICE3_EMOTIONS)[number];
2285
+ export type TtsCosyvoice3Instruct = z.infer<typeof ttsCosyvoice3InstructSchema>;
1314
2286
  export type TtsChatterboxLoadConfig = z.infer<typeof ttsChatterboxLoadConfigSchema>;
1315
2287
  export type TtsSupertonicLoadConfig = z.infer<typeof ttsSupertonicLoadConfigSchema>;
1316
2288
  export type TtsParlerLoadConfig = z.infer<typeof ttsParlerLoadConfigSchema>;
2289
+ export type TtsCosyvoice3LoadConfig = z.infer<typeof ttsCosyvoice3LoadConfigSchema>;
2290
+ export type TtsAudio8LoadConfig = z.infer<typeof ttsAudio8LoadConfigSchema>;
1317
2291
  export type TtsLoadConfig = z.infer<typeof ttsLoadConfigSchema>;
1318
2292
  /** @deprecated Use {@link TtsChatterboxLoadConfig} */
1319
2293
  export type TtsChatterboxConfig = TtsChatterboxLoadConfig;
@@ -1322,6 +2296,8 @@ export type TtsSupertonicConfig = TtsSupertonicLoadConfig;
1322
2296
  export type TtsChatterboxRuntimeConfig = z.infer<typeof ttsChatterboxRuntimeConfigSchema>;
1323
2297
  export type TtsSupertonicRuntimeConfig = z.infer<typeof ttsSupertonicRuntimeConfigSchema>;
1324
2298
  export type TtsParlerRuntimeConfig = z.infer<typeof ttsParlerRuntimeConfigSchema>;
2299
+ export type TtsCosyvoice3RuntimeConfig = z.infer<typeof ttsCosyvoice3RuntimeConfigSchema>;
2300
+ export type TtsAudio8RuntimeConfig = z.infer<typeof ttsAudio8RuntimeConfigSchema>;
1325
2301
  export type TtsRuntimeConfig = z.infer<typeof ttsRuntimeConfigSchema>;
1326
2302
  export type TtsConfig = z.infer<typeof ttsConfigSchema>;
1327
2303
  export type TtsClientParamsInput = z.input<typeof ttsClientParamsSchema>;