Audio generation models

Audio generation models turn text into audio bytes. Use them for narration, accessibility, voice responses, and media workflows.

Create an audio model

import { TTS_1, OpenAIClient } from '@anvia/openai'

const openai = new OpenAIClient({
  apiKey: process.env.OPENAI_API_KEY,
})

export const audioModel = openai.audioGenerationModel(TTS_1)

OpenAI and Grok currently provide audio-generation adapters.

Generate audio

import { writeFile } from 'node:fs/promises'
import { audioGenerationRequest } from '@anvia/core/audio-generation'

const response = await audioGenerationRequest(audioModel)
  .text('Your incident summary is ready.')
  .voice('alloy')
  .speed(1)
  .additionalParams({ response_format: 'mp3' })
  .send()

await writeFile('incident-summary.mp3', response.audio)

The response contains audio bytes, an optional media type, and the raw provider response.

Production boundary

Validate text length, voice, speed, and format before generation. Store generated bytes in media storage instead of agent memory or event logs, and use a queue for long scripts or bulk work.