Class PiperNativeTTS

Native Piper TTS voice with persistent ONNX session (mirrors piper1-gpl voice.py PiperVoice).

Index

Methods

getConfig getModelPath phonemize synthesize synthesizeChunks load

Methods

getConfig

getConfig(): PiperConfig
Returns the resolved Piper voice configuration.

Returns PiperConfig
Sample rate, phoneme type, id map, and defaults parsed from the .onnx.json file.
- Defined in native/voice.ts:231

getModelPath

getModelPath(): string
Returns the resolved local .onnx model path (downloaded first when the voice came from a URL or catalog reference).

Returns string
Absolute model path.
- Defined in native/voice.ts:240

phonemize

phonemize(text: string): Promise<string[][]>
Converts text to phonemes with the voice's own phonemizer (espeak, text, hebrew, lithuanian, pinyin, thai, or japanese; Arabic ar voices are diacritized first unless disabled).
Parameters
- text: string
  Input text to phonemize.
Returns Promise<string[][]>
Phoneme arrays grouped by sentence.
- Defined in native/voice.ts:250

synthesize

synthesize(
text: string,
options?: NativeSynthesizeOptions,
): Promise<{ audio: Buffer; sampleRate: number; text: string }>
Synthesizes text and returns the full audio: sentence chunks concatenated with sentenceSilence gaps, then encoded to outputFormat (wav by default) and optionally written to outputFile.
Parameters
- text: string
  Input text to synthesize.
- options: NativeSynthesizeOptions = {}
  Per-call synthesis parameters, output format, and optional output file.
Returns Promise<{ audio: Buffer; sampleRate: number; text: string }>
Final audio with its sample rate and source text.

Throws
When text is empty or no audio is produced.
- Defined in native/voice.ts:474

synthesizeChunks

synthesizeChunks(
text: string,
options?: NativeSynthesizeOptions,
): AsyncGenerator<NativeAudioChunk>
Synthesizes text sentence by sentence, yielding raw PCM per sentence as it is produced.
Parameters
- text: string
  Input text to synthesize.
- options: NativeSynthesizeOptions = {}
  Per-call synthesis parameters (scales, speaker, volume).
Returns AsyncGenerator<NativeAudioChunk>
Per-sentence { sampleRate, phonemes, phonemeIds, pcm16 } chunks.

Throws
When text is empty.
- Defined in native/voice.ts:415

`Static`load

load(options: NativeTtsOptions): Promise<PiperNativeTTS>
Loads a voice from a local file, a direct URL, or a catalog reference.

Resolution order: explicit modelPath, then modelUrl (downloaded once into modelsDir and cached), then model (rhasspy id, hub:<slug>, piperhub.org page URL, or bare PiperHub slug). The .onnx.json config defaults to the model sibling (<model>.json / <modelUrl>.json).
Parameters
- options: NativeTtsOptions
  Model source, session reuse, and phonemizer data locations.
Returns Promise<PiperNativeTTS>
A voice with a warm persistent ONNX session.

Throws
When no usable source is provided, files are missing, or the config/phoneme type is unsupported.
- Defined in native/voice.ts:180