@mayurajs/voice-deepgram

Deepgram speech-to-text and text-to-speech for Mayura's voice registry, over Deepgram's HTTP API.

npm install mayura @mayurajs/voice-deepgram

import { createVoices } from 'mayura/voice';
import { deepgramVoice } from '@mayurajs/voice-deepgram';

const voices = createVoices({ providers: [deepgramVoice({ apiKey })], prices: 'catalog', maxCallCostMicros: 50_000 });
const speech = await voices.speaker('deepgram/aura-2').speak({ text: 'Hello.', voice: 'thalia-en' });

• Ids are deepgram/<model>, granted as voice:<id>. Speech ids name a model family that the voice completes: deepgram/aura-2 with voice thalia-en is Deepgram's aura-2-thalia-en. The catalog lists Nova-3 (at its multilingual rate), Whisper Large, Aura-2 and Aura at Deepgram's pay-as-you-go prices on its date.
• No dependencies: requests go through fetch. Nothing is read from the environment.
• Every call is bounded before it is sent. Passes Mayura's voice adapter conformance suite.

See the voice guide. Apache-2.0.