Offline Audio Transcriber

Offline Audio Transcriber is an audio transcription app that turns audio to text on your phone: MP3 to text, voice notes to text, recorded speech to text. The audio is never uploaded: no account, no server, and no internet needed once the speech model is on your phone.

Transcribe audio to text free: an audio to text converter and mp3 to text converter that turns speech into text on your phone.

WHAT YOU CAN DO
- Record with the built-in recorder, or import MP3, M4A, WAV, OGG, MP4 and other audio and video files to transcribe audio to text, including WhatsApp voice messages.
- Read the transcript with timestamps while the audio plays. Tap a word to jump to that moment.
- Fix mistakes in the text, search it, and sort recordings into folders.
- Export as TXT, SRT or VTT subtitles, copy the text, or share it.

SPEED
The speech model runs on the phone's processor. On a Samsung Galaxy S23 Ultra, a one-hour recording took about 8 minutes (our own test; older and cheaper phones are slower). Long jobs keep running in the background with a progress notification.

LANGUAGES
The main model understands 24 European languages: Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Slovak, Slovenian, Spanish, Swedish and Ukrainian. An optional second model (Whisper Base) adds more languages and is slower.

WHAT IT COSTS
- The first 2 minutes of every recording are free.
- Watch a short ad to add 3 minutes.
- Or buy minute packs that never expire, or Pro for unlimited minutes (monthly, yearly or one-time).
- The speech model is a one-time download of about 671 MB. The app shows the size before it starts; use Wi-Fi.

PRIVACY
No account. Recordings and transcripts stay on your phone. The app goes online only to download the speech model, to show ads and for purchases through Google Play.

GOOD TO KNOW
Accuracy depends on audio quality, accents and background noise. Check important transcripts before you rely on them. The main speech model, NVIDIA Parakeet TDT 0.6B v3, is licensed under CC BY 4.0.

Recent changes:
Watching a short ad now gives you 5 extra minutes of transcription (was 3). Clearer message before the ad.