Whisperscribe transcribes your audio files to text directly on your Mac and organizes them into a searchable library. This guide takes you from first launch to your first transcription, and resolves the most common issues.
Requirements
Whisperscribe requires macOS 14 (Sonoma) or later. All processing happens on your device, using the Apple Neural Engine.
Because the models run locally, we recommend at least 8 GB of RAM and free disk space: ~0.6 to 3 GB depending on the transcription model you pick, plus ~1.5 GB for the summary model.
Getting started: download a model
When you open the app for the first time you'll see the model download. You need at least one transcription model to begin; the summary model is optional.
You can manage downloads anytime from Settings (⌘,). The model download is the only network connection the app makes; once downloaded, everything works offline.
- Compact — multilingual, ~0.6 GB. The lightest and fastest.
- Balanced — multilingual, ~1.0 GB. Default, a good balance of speed and accuracy.
- Maximum accuracy — ~3.0 GB. The most precise, at the cost of speed and disk.
- Summary model — a local language model (~1.5 GB) to generate summaries and tags.
Import audio
Add audio with the ⊕ button in the toolbar or by dragging files directly onto the transcript list.
When importing you can set the suggested language: Automatic, English, or Spanish. If a file is already in your library, the app detects it and lets you import it anyway or skip it.
- Supported formats: MP3, M4A, WAV, and AIFF.
- You can import several files at once.
- Audio is copied into an app-managed folder on your Mac.
Transcribe
If you have a transcription model installed, transcription starts automatically when you import. A progress bar shows each file's status: Pending → Transcribing → Done (or Error).
If you haven't installed a model yet, files queue up and are transcribed as soon as one is available.
Summaries and tags
When a transcription finishes, Whisperscribe generates a summary and suggested tags locally, using the language model. Everything happens on your Mac, with nothing sent to the cloud.
You can regenerate the summary anytime. If you haven't installed the summary model, this step queues until you download it.
Organize your library
Keep your transcriptions tidy with projects (and subprojects), tags, and full-text search across all your content.
- Drag transcriptions onto a project in the sidebar to reorganize them.
- Rename a transcription inline: hover over the title and click the pencil icon.
- Right-click transcriptions or projects to open the context menu (rename, delete, etc.).
- Use the search field to find any word within your transcriptions.
Play and navigate
The player stays pinned at the top of the detail pane, with a scrubber to move through the audio while you read.
Click any timestamp (for example [00:45]) inside the transcript to jump to that moment. The active segment is highlighted and the view auto-scrolls during playback.
Privacy
Whisperscribe is local and private by design. Your audio, transcriptions, and summaries never leave your Mac.
The only network connection is the initial model download. After that, the app works completely offline.
Troubleshooting
A model shows as “Corrupted”.
This usually happens if the download was interrupted. Go back to Settings (⌘,) and retry the model download.
“The audio file is missing” or “was moved or deleted”.
The original file is no longer where the app expected it. Re-import the file to play or transcribe it again.
Transcription or the summary won't start.
Make sure you have the matching model installed: a transcription model to transcribe, and the language model for summaries. Without it, the work stays queued.
I'm running out of disk space.
Open the Storage tab in Settings to see how much the models, audio files, and database take up, and free space from there.
Need more help?
If something wasn't clear or you ran into a problem that isn't here, write to us at [email protected] and we'll gladly help. You can also check the Support page.