@kpanks/voice

Text-to-speech for Praetor charters. Default backend is Kokoro 82M (Apache 2.0, ONNX, lazy-loaded — no API key, no quota, no network egress). Azure Speech, Edge TTS, ElevenLabs, OpenAI TTS, fal-tts, and OpenVoice v2 ship as opt-in adapters tagged with licenseFamily so charters can refuse non-permissive backends.

Install

npm install @kpanks/voice
# Optional peer for the Kokoro adapter:
npm install kokoro-js

Usage

import { PraetorVoice, KokoroAdapter, AzureSpeechAdapter } from "@kpanks/voice";

const voice = new PraetorVoice();
voice.attach("kokoro", new KokoroAdapter());
voice.attach("azure-speech", new AzureSpeechAdapter({ key, region }));

const out = await voice.synthesize({ text: "hello", voice: "af_bella" });
// out.audioBuffer, out.mime, out.backend, out.licenseFamily

License families

| Family | Adapters | |---|---| | apache_or_mit | Kokoro, Piper, OpenVoice v2 (safe-default) | | proprietary | Azure Speech, ElevenLabs, OpenAI TTS, fal-tts | | restricted | non-commercial / Llama-derivative / watermarked |

Pass allowedLicenseFamilies on PraetorVoice to whitelist.

License

Apache 2.0. Kokoro itself is Apache 2.0.