@nhtio/adk/batteries/tts/_shared

Structural contracts shared across every TTS (text-to-speech) engine — the model-backed transformers.js engine and the OS-native CLI engine both build on these.

Remarks

This module owns the shared TTS option base (BaseTtsAdapterOptions) and the shared per-call shape (SynthesizeOptions), plus the audio-result contract (RawAudioLike). It lives in _shared — NOT inside any one engine — deliberately: the native engine is model-less while the transformers.js engine requires a model, so neither engine "owns" the common base. A future non-transformers engine can extend this base without importing a transformer-named subpath.

model is intentionally absent from BaseTtsAdapterOptions: the OS-native engine synthesizes through say/espeak-ng/PowerShell and has no model concept, so each model-backed engine adds its own required model field rather than the base forcing one. Do not "fix" this by hoisting model into the base — it is a deliberate divergence from the embeddings/generation domains (whose bases require model).

Interfaces

Interface	Description
BaseTtsAdapterOptions	Options shared by every TTS engine, regardless of backend.
RawAudioLike	The minimal structural slice of transformers.js's RawAudio the TTS engine consumes: a single toBlob() that encodes the synthesized samples to a WAV Blob.
SynthesizeOptions	Per-call options accepted by every engine's synthesize. Engine-specific per-call option interfaces extend this with their own knobs.

Type Aliases

Type Alias	Description
TtsSynthesisResult	The synthesized-audio result every TTS engine resolves to — an alias of the shared GeneratedMediaOutput descriptor, always kind: 'audio' in practice.

References

GeneratedMediaOutput

Re-exports GeneratedMediaOutput