Audio in.
Text out. No upload.
Drop in a file or share it from any app and ThinkScribe transcribes it on your own device, with speakers told apart and every word synced to the audio. Any length, any of 90+ languages, no upload.
One purchase covers Mac, iPhone, iPad & Apple Watch.
Three steps, no upload.
- 1
Add the audio
Share a file from Files, Voice Memos, WhatsApp or Telegram, or drag it into the Mac app. MP3, M4A, WAV, FLAC and OGG all work.
- 2
It transcribes on your device
Pick Apple’s speech models, Whisper or Parakeet. The language is detected automatically, or set it yourself from 28 languages.
- 3
Read, search, export
Tap any word to hear that moment. Export TXT, SRT, VTT, JSON, Markdown, PDF or DOCX, or send a summary to Apple Notes.
Everything a transcription service does, without the service.
Any file, any length
A three-hour recording costs the same as a three-minute one. There is no per-minute meter and no monthly quota.
Speaker labels
Voices are told apart automatically and remembered across recordings. Name a speaker once and the name sticks.
90+ languages
Automatic language detection, plus a picker with 28 explicitly supported languages for when you want to pin it.
Word-synced playback
Every word is linked to its timestamp. Click a word to jump the audio there and check it against the recording.
Summaries and answers
Summarize, pull action items or ask a question. The AI answers with citations that jump to the audio.
Seven export formats
TXT, SRT, VTT, JSON, Markdown, PDF and DOCX, so the transcript fits whatever comes next.
Questions, answered.
How do I transcribe an audio file to text on a Mac?
Open ThinkScribe, drag the file in (or use Import), and the transcript appears after a moment. On iPhone, share the file to ThinkScribe from any app. The transcription runs on the device itself, with no upload.
Which audio formats can I import?
MP3, M4A, WAV, FLAC and OGG, shared from any app or picked from Files. Voice memos and WhatsApp voice notes work the same way.
Is there a length limit?
No. Transcription is unlimited on every plan, and a long recording is processed in the background while you keep using your device.
How accurate is it?
ThinkScribe runs the same open speech models used in research and production (Whisper, Parakeet, Apple’s models). Clear audio gives near-verbatim text; you can switch engines per recording when one suits the material better.