Turn recordings and videos into text, on your Mac
Meeting recordings, lectures, interviews, and videos become timestamped text you can copy or export. The audio is recognized on your Mac and never uploaded.
Updated 2026-11-05

How to transcribe a file
- Click the Tidytalk icon in the menu bar and choose "Transcribe File or Link…". This opens the Transcribe page in the Tidytalk window. Use the Files tab.
- Drop an audio or video file onto the page, or click the drop area to choose one. MP3, M4A, WAV, MP4, MOV and other common formats work, up to 3 hours long.
- Tidytalk splits the recording at pauses and recognizes it sentence by sentence. Each sentence shows when it was said.
- Each transcription is kept as a card. Open it to play it back or fix a line, and use the copy and export buttons at the top to get the text as plain text, SRT subtitles, or Markdown.
Transcribe a video link
On the Links tab, paste a link to a single video from YouTube, Bilibili, or another site, and press Return or click the arrow button. The first time, Tidytalk downloads yt-dlp, an open-source download tool (about 35 MB), from its official GitHub release and checks it against the published checksum. After that, only the video's audio and cover image are downloaded. They stay on your Mac: the card shows the cover and length, and you can play or save the audio, or delete it and keep the text.
Use links only for content you have the right to use. Videos that need you to sign in, playlists, and live streams aren't supported.
Which speech model to use
File transcription uses the model you choose in Settings › Speech Model. All of them run on your Mac.
| Model | Best for |
|---|---|
| SenseVoice (default) | Fast transcripts in Chinese, English, Japanese, Korean, or Cantonese, including mixed Chinese and English. |
| Qwen3-ASR | The most accurate results and the best punctuation, in 30 languages and 22 Chinese dialects. It also recognizes names and terms from your vocabulary. |
| Whisper Turbo | Less common languages; it covers about 99. Use another model for Chinese. |
Export formats
- Plain text (.txt): paragraphs split at longer pauses, ready for notes or an AI summary.
- Subtitles (.srt): one cue per sentence with start and end times, for video editors and players.
- Markdown (.md): the title and each sentence with its timestamp, for docs and wikis.
Tips and limits
- Clear audio with one person speaking at a time gives the best results. Background music and people talking over each other lower accuracy.
- Tidytalk doesn't label who is speaking.
- Your replacements and Chinese–English spacing are applied to the transcript, just like dictation.
- For a meeting that's happening right now, use live captions instead: press ⌃⌥C to caption what your Mac is playing as it plays.
Privacy
Recognition happens on your Mac, so recordings are never uploaded. Transcripts stay in the window until you copy or export them, and nothing is saved anywhere else. For links, the only network traffic is downloading yt-dlp once and the video's audio.