VocalLab Speech to Text turns spoken audio into accurate text — across 30 languages, from a short voice note to a six-hour recording.

Transcribe audio

1. Open Speech to Text in the Studio sidebar, under Create.
2. Drag in an audio or video file, or click Record audio to use your microphone.
3. Leave the language on Auto-detect — set it yourself only if a previous result came out in the wrong language.
4. Click Transcribe.

Supported formats: MP3, WAV, M4A, OGG, FLAC and WebM, plus MP4 and MOV video — the audio is extracted for you. How long a file may be, and how large, depends on your plan.

Long recordings

Short files transcribe in seconds, right where you are. Anything longer is split at natural pauses, transcribed in parts and stitched back together, which is what makes hours-long recordings possible at all.

Long jobs run in the background, so you can leave the page and come back — the transcript is waiting when you return. While one is running you'll see the stage it has reached and how many parts are done.

• Cancel at any time. Points for the parts that never ran are refunded.
• If a part can't be read, you get the rest of the transcript and are charged only for what worked.

After transcribing

• Copy the transcript or download it as a .txt file.
• Click Use in Text to Speech to send it straight to the TTS editor.
• Every transcription is saved to your History (the Speech to Text tab).

Languages

The language is detected automatically across 30 languages — including English, Spanish, French, German, Italian, Portuguese, Dutch, Polish, Russian, Arabic, Hindi, Chinese, Japanese and Korean.

You don't normally need to pick one. It's worth setting the language by hand when detection has something to get wrong: a very short clip, a heavy accent, music under the speech, or two languages that share a script.

For the best accuracy, use a clean recording with minimal background noise and one speaker at a time.

Plans and limits

Cost

Transcription costs points based on audio length — 15 points per minute, whatever the file's length. You only pay for the audio you transcribe, and a job you cancel only costs you the parts that had already run.