Anvia

Appearance

Speech and transcription ​

Type: Pattern

Outcome ​

Generate speech from text, save the audio, then transcribe it through Anvia's provider-neutral media requests. Use this pair to learn both contracts; a real application can adopt either half alone.

Prerequisites ​

  • Node.js 22 or newer
  • @anvia/core, @anvia/openai, tsx, and a server-side OPENAI_API_KEY
  • Permission to process the source voice or audio

Implementation ​

ts
import { readFile, writeFile } from 'node:fs/promises'
import { audioGenerationRequest } from '@anvia/core/audio-generation'
import { transcriptionRequest } from '@anvia/core/transcription'
import { OpenAIClient } from '@anvia/openai'

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

const speech = await audioGenerationRequest(client.audioGenerationModel())
  .text('Anvia provides provider-neutral audio and transcription contracts.')
  .voice('alloy')
  .speed(1)
  .additionalParams({ response_format: 'mp3' })
  .send()

await writeFile('speech.mp3', speech.audio)

const transcript = await transcriptionRequest(client.transcriptionModel())
  .data(await readFile('speech.mp3'))
  .filename('speech.mp3')
  .prompt('Transcribe the audio exactly.')
  .temperature(0)
  .send()

console.log(transcript.text)

Run and expected behavior ​

Run pnpm tsx audio.ts. It writes speech.mp3, then prints a transcript close to the source text. Exact punctuation can vary. The two calls are billed and fail independently.

Boundaries ​

Obtain consent and follow local recording and biometric-voice laws. Limit upload duration, size, codec, and type; scan files; and never assume a transcript is exact. Do not use generated speech to impersonate a person or conceal that audio is synthetic.

In production, move long media work to durable jobs, store protected inputs and outputs with retention and deletion, redact sensitive transcript content, expose job status, and evaluate noisy, multilingual, accented, and domain-specific audio.

Source and extensions ​

Run the OpenAI audio cookbook. Next, transcribe an uploaded file, add timestamps at the provider boundary if supported, or stream job progress to a UI.

Built for Anvia.