What On-Device Transcription Actually Means

What On-Device Transcription Actually Means

Understand where audio, captions, models, and language resources live in a local-first workflow.

Yes. After the required Apple language resources or local Whisper model are downloaded once, audio capture, recognition, and translation run on the Mac. Internet may be needed only for the first resource download.

Processing stays on the Mac

Alison processes captured audio frames in memory and does not send them to an app-operated transcription backend.

Downloads are different from uploads

Apple language resources and optional Whisper models may need a one-time download before recognition or translation runs locally.

Text history is still stored data

Caption text and translations remain in Application Support until you delete them; raw audio is not added to history.

Offline file transcription

Import local audio or video, edit the transcript and translate it on your Mac after the required resources are installed.

Does it automatically summarize my recordings with AI?

Alison creates editable transcripts and exports text, including Markdown. Summarization in a separate AI tool is your choice; sending text to a cloud tool changes where that text is processed.

Can I caption my voice and the other side of a call?

Yes. Choose microphone and system audio together for separate source labels and independent recognition and translation languages. Use headphones to reduce echo. This does not identify every remote speaker.

Choose the next guide for your task

Use the related guides below to continue with the task you actually need: following live speech, translating it, preparing a file transcript, checking privacy or fixing audio permissions. Each guide explains its own setup and limits. Check your actual app, audio route and supported languages before a long session.