Apple Music Hear your music the way it was mastered. Get up to 3 months of Apple Music free, with lossless audio and Dolby Atmos. Try it free Ad · New subscribers only
SoundBench فارسی
Voice & recording

Audio to Text Converter

Turn speech in any audio file or recording into text with AI. It runs on your own device, supports more than 90 languages and exports plain text or SRT and VTT subtitles.

About the Audio to Text

Audio to Text uses OpenAI's open Whisper speech recognition model, downloaded once and run on your own device. Unlike most transcription sites, your audio is never uploaded, which makes it a good fit for interviews, meetings, lectures and other private recordings.

Choose the language or let it be detected, pick Fast, Balanced or Accurate, and watch the text appear as it is transcribed. Copy the result, download it as TXT, or save SRT and VTT subtitle files with timestamps.

How to use the Audio to Text

  1. Drop an audio file onto the page, or record with your microphone.
  2. Choose the language and accuracy, then click Transcribe. The model downloads the first time.
  3. Copy the text or download it as TXT, SRT or VTT.

Frequently asked questions

Is the transcription really free?

Yes. There are no minutes, credits or accounts. The AI runs on your device, so there is nothing for us to charge for.

Which languages are supported?

Whisper understands more than 90 languages, including English, Persian, Arabic, Turkish, Spanish, French, German, Russian, Hindi, Chinese and Japanese. Use Accurate for languages other than English.

How long does it take?

It depends on your device and the accuracy setting. On a modern computer the Fast model handles an hour of audio in a few minutes; Accurate takes several times longer. Phones are slower.

Can I make subtitles for a video?

Yes. Extract the audio with Video to MP3, transcribe it here and download the SRT file, which most video players and editors can load.

Can it write down the lyrics of a song?

Yes, with some care. Tick "This is a song" so the vocals are separated from the music first, and use the Best model. Sung words are harder to recognize than speech, so expect some mistakes, especially with heavy effects or fast singing.

Related tools