Skip to content

Speech translation

Hear one language. Read another.

Upload a recording and get back what was said, written in any of 19 languages. The spoken language is detected from the audio.

Open the translatorRead the docs

Play a recording and read what the model returned for it.

Translate your own fileUpload a recording in the studio.

Restaurant booking

French to English

Original, French

Bonjour, je voudrais réserver une table pour quatre personnes vendredi soir, vers vingt heures. Est-ce qu'il serait possible d'avoir une table en terrasse ?

Translation, English

Hello, I would like to reserve a table for four people for Friday evening around eight o'clock. Would it be possible to have a table on the terrace?

Real, unedited output from symphony.

How it works

From a recording to readable text in three steps.

  1. 01

    Upload a recording

    Drop in an audio file: MP3, WAV, M4A, FLAC, OGG, and other common formats.

  2. 02

    Pick a language

    Choose any of 19 languages to read it in. English is the default.

  3. 03

    Read the translation

    Copy it, download it as text, or open it again later from your file list.

For developers

One request, 19 languages.

The OpenAI translations endpoint, plus a target_language field that sets the language the text comes back in. Point your existing client at Sprag and add a target language.

translate.py
from openai import OpenAI

client = OpenAI(
    api_key="sk-sprag-...",
    base_url="https://api.sprag.ai/v1",
)

with open("call.mp3", "rb") as f:
    translation = client.audio.translations.create(
        model="symphony",
        file=f,
        extra_body={"target_language": "de"},
    )

print(translation.text)
  1. Upload the recordingSend the file as multipart form data to /v1/audio/translations.
  2. Name the language you wantSet target_language to an ISO-639 tag. Leave it out and the text comes back in English.
  3. Read the textThe translation comes back as text in one JSON response.
Get an API keyRead the docs

Audio in, text out

One recording, one translated text.

Language detected

Name the spoken language only if you know it.

Drop-in OpenAI

The translations endpoint your client already has.

Any common format

MP3, WAV, M4A, FLAC, OGG, and more.

Sprag Studio

Where translation fits.

Speech-to-TextA transcript in the language that was spoken, with word timings and speakers.Text-to-SpeechRead the translation aloud in a preset, designed, or cloned voice.

Frequently asked questions

Make your first request.

Get startedAPI reference