Skip to content

TextToSpeechModel

Defined in: packages/runntime/src/zoo/tasks/audio/textToSpeech.ts:53

What to load. Every field is optional; a missing path comes from models.textToSpeech.REVERB.DEFAULT.

readonly optional codecPath?: ModelPath

Defined in: packages/runntime/src/zoo/tasks/audio/textToSpeech.ts:58

The Higgs Audio v2 tokenizer checkpoint. createTextToSpeech reads its decoder, createVoiceMaker its encoder.


readonly optional dtype?: "f32" | "f16"

Defined in: packages/runntime/src/zoo/tasks/audio/textToSpeech.ts:64

Precision on the GPU of the math and the float weights. Needs a device with shader-f16 for f16. Default: f16 where the device has it, f32 elsewhere. Either loads any weight file.


readonly optional modelPath?: ModelPath

Defined in: packages/runntime/src/zoo/tasks/audio/textToSpeech.ts:55

The Reverb checkpoint. The weight format is read from the file.


readonly optional tokenizerPath?: string

Defined in: packages/runntime/src/zoo/tasks/audio/textToSpeech.ts:60

The Hugging Face tokenizer.json.


readonly optional verbalize?: (text, language) => string

Defined in: packages/runntime/src/zoo/tasks/audio/textToSpeech.ts:69

Rewrites the text and the voice’s transcript as they are read aloud, for example numbers as words: (text, language) => new Verbalizer(language).process(text) from verbalis. Default: none.

string

"en" | "de" | "fr" | "es" | "it" | "pl" | "pt"

string