# Speech Translation

Purpose: Agent-readable overview of Sprag speech translation.
Last updated: 2026-10-12

Sprag speech translation takes a finished audio recording and returns what was said as text in another language. The spoken language is detected from the audio unless you name it in source_languages. It runs on Symphony, Sprag's multimodal model.

## How it works

1. **Upload a recording.** MP3, WAV, M4A, FLAC, OGG, and other common audio formats.
2. **Pick a language.** Any of the 19 target languages below. English is the default.
3. **Read the translation.** The text comes back in one JSON response.

## Target languages

Arabic (ar), Cantonese (yue), Chinese (zh), Dutch (nl), English (en), French (fr), German (de), Hindi (hi), Indonesian (id), Italian (it), Japanese (ja), Korean (ko), Malay (ms), Portuguese (pt), Russian (ru), Spanish (es), Turkish (tr), Urdu (ur), Vietnamese (vi).

## API

- Endpoint: POST /v1/audio/translations (OpenAI-compatible), multipart/form-data.
- Model: symphony.
- Fields: file (required), model (required), target_language (two- or three-letter ISO-639 tag, English when omitted), source_languages (ISO-639 tag of the spoken language, optional and repeatable), response_format (json only).
- Output: {"text": "...", "usage": {...}}.
- OpenAI Python SDK: client.audio.translations.create(model="symphony", file=f, extra_body={"target_language": "de"}).
- Docs: https://sprag.ai/docs/stt/translation.md

## Pricing

Billed at Symphony's rates for the audio you send and the text returned. An hour of audio comes to about $0.09. See https://sprag.ai/pricing.md.

Product page: https://sprag.ai/translation