Voice Transcription

Transcribe captured voice memos into text.

Two Backends

Backend	Where it runs	Pros	Cons
Local Whisper	Your machine	Private, free, offline	Larger models = larger downloads
OpenAI Whisper API	OpenAI cloud	Faster on small devices, latest models	Requires API key, content leaves your device

Pick a backend in Settings → AI → Voice Transcription.

Local Whisper

memrynote uses Whisper Small as the default local model — a balance of accuracy and speed.

Setup

1. Settings → AI → Voice Transcription → Provider: Local
2. Click Download for the model
3. Wait for status: Loaded

Model files are stored locally in the app data directory. Disk usage is shown next to each option.

Local transcription starts a native helper process while Whisper runs. After a transcription request finishes and no new voice work starts, memrynote shuts that helper down after a short idle period so the model can release memory.

Languages

Whisper Small handles dozens of languages well. For unusual languages, larger Whisper models tend to do better — see the open-source Whisper docs for tradeoffs.

OpenAI Whisper API

For machines that struggle with local Whisper or when you want maximum accuracy:

1. Settings → AI → Voice Transcription → Provider: OpenAI
2. Paste your API key
3. The key is stored locally in the vault, encrypted at rest

Each transcription call counts against your OpenAI quota.

Memo Naming

Voice memos can be renamed automatically after transcription. In Settings -> AI -> Voice Transcription -> Memo names, choose:

• Transcript title - uses the first sentence of the transcript, trimmed to a short title
• Timestamp - keeps a predictable Voice memo YYYY-MM-DD HH:mm name
• Do not rename - leaves the original capture title unchanged

Renaming only applies while the item still has its generated voice memo title; edited titles are preserved.

How Voice Memos Are Captured

The Inbox supports voice capture from the header and inline quick capture surfaces. Recorded audio is queued for transcription.

Transcribed items appear in the inbox under the voice content type filter. The original audio is preserved as an attachment.

When a transcribed voice memo is filed into a note, task, or project, the audio attachment keeps its transcript with the file so the player and transcript stay together.

If voice setup is incomplete, memrynote opens AI settings instead of starting a recording.

What's Sent Where

Provider	What leaves your device
Local Whisper	Nothing
OpenAI Whisper	Audio file → OpenAI API; transcript returns

memrynote does not forward audio through the sync server.

Transcription Quality

For clear speech in supported languages, Whisper produces near-broadcast quality transcripts. Background noise, multiple speakers, and technical jargon hurt accuracy. Editing the resulting text is normal — memrynote treats the transcript as a starting point.

Disabling

Disabling Enable in the AI panel halts new transcriptions. Existing inbox voice items stay; their transcripts (if any) remain.

See Also

• Inbox
• Provider Setup