Build voice into your apps and agents.

Dictation and transcription. One key for both.

Dictation

For one speaker, up to 10 minutes. Returns finished text with the user’s language, dictionary, replacements, and instructions.

Transcription with Avalon

For any audio, up to four hours. Works with the OpenAI SDKs: change the base URL and model. About Avalon

In production. The route or header for each.

  • Retry without a second charge.

    Idempotency-Key

  • Run long audio in the background.

    POST /v1/audio/transcription-jobs

  • Read a user’s past dictations.

    GET /v1/sessions

  • Use each user’s dictionary and settings.

    GET /v1/customizations

  • Act for a member of your organization.

    Aqua-User

  • Check that transcription works.

    GET /v1/health/functional

97.4% of technical terms. Right the first time.

Key-term recognition on AISpeak-10, our benchmark of technical speech. See the benchmarks

  • Avalon97.4%
  • ElevenLabs Scribe78.8%
  • Whisper Large v365.1%
  • Voxtral Mini 3B59.5%
  • NVIDIA Canary 1B51.5%

Try it free, then pay by the hour.

Avalon transcription

$0.39/ hour of audio

  • Any audio, up to four hours per file
  • Timestamps, subtitles, and speaker labels with transcription jobs
  • Works with the OpenAI SDKs

Aqua dictation

$0.49/ hour of audio

  • One speaker, up to 10 minutes
  • Finished text with each user’s dictionary, replacements, and instructions
  • Edit Mode at the same rate

Your first $1.00 of usage is free. Both are billed per second, with a 10-second minimum. Failed requests and idempotent replays are never billed, and there are no seat fees.

Frequently asked questions

What is the difference between the Aqua API and the Avalon API?

Avalon is Aqua's speech-to-text model, served on the OpenAI-compatible transcription routes. The Aqua API includes those routes and adds dictation, voice editing, stored sessions, customizations, settings, and MCP.

Is the Aqua API OpenAI-compatible?

Only the transcription routes. The official OpenAI SDKs work with them after you change the base URL and model name. Dictation, sessions, customizations, and settings use Aqua's own request and response shapes.

Which audio files can I transcribe?

Avalon accepts flac, mp3, mp4, mpeg, mpga, m4a, ogg, opus, wav, and webm files, up to 256 MiB and four hours of decoded audio.

Can I get subtitles or timestamps?

Yes. Response formats include json, text, srt, vtt, and verbose_json. Segment timestamps are supported with verbose_json.

Can transcription run in the background?

Yes. Use the transcription-jobs endpoint to submit an asynchronous job, poll its status, and retrieve its result. Synchronous transcription is also supported.

How do I get an API key?

Create a key in the Aqua API dashboard. One key covers dictation, sessions, customizations, settings, MCP, and transcription. Your account activates with the first key, and the first $1.00 of usage is free, so the first request needs no card. Each account can hold up to 30 keys.

Which languages does the API support?

Dictation supports 50 languages, with the best quality in Dutch, English, French, German, Italian, Japanese, Korean, Mandarin, Portuguese, Russian, Spanish, Turkish, and Ukrainian. Transcription detects the language, or takes any ISO 639-1 code as a hint.

Can I send a meeting to the dictation API?

No. Dictation is for one speaker and can drop other voices. Use transcription jobs, which can label speakers.

What does Aqua store from an API dictation?

It depends on the user's privacy and zero-data-retention settings. With retention on, Aqua stores the dictation in history and returns its session ID. With retention off, Aqua returns the text and does not store it.

Can my AI agent use Aqua?

Yes. Aqua is also a remote MCP server. Claude, ChatGPT, Codex, and other clients connect over OAuth to read sessions and transcripts, send feedback, and manage dictionary terms and settings.