Transcribing audio into a document

What this does

Task menu ▸ Transcriptor

Two routes to written text: you record directly, or you upload an existing audio file. The result can be inserted into your document.

The window

• Microphone – which input device is used. If none is detected, recording is disabled and the window says so.
• Language – the language of the recording.
• Record / Stop – the audio is only sent for processing when you press Stop.
• Upload – for existing files: .wav, .mp3, .m4a, .mp4, .aac, .ogg, .opus, .webm, .flac, .wma.
• Insert format – Text only or With timestamps. Timestamps pay off for meetings, when you want to find passages again later.
• Insert into document – puts the transcript at the cursor position.

The limit: 25 MB, about 13 minutes

That is the number that matters. While recording, the status line shows how much of the budget is left. When it runs out, the recording stops by itself – what was recorded up to then is transcribed, the rest was never recorded.

For a longer meeting that means: record in sections. Plan to stop and restart every ten minutes rather than relying on the automatic limit.

The same limit applies to uploads. A file that is too large is rejected, with its size given – split it beforehand.

What to check

A transcript is an approximation. Experience says these go wrong: proper names, company names, terms of art, figures, and anything said over the top of somebody else.

If the transcript is to serve as minutes, read it against your own recollection before passing it on.

When it does not work

• "No LiteLLM transcription models are available." – no transcription model is set up for your account. Your administrator does that.
• "No audio input device detected" – no microphone connected, or not released by Windows. Uploading still works.
• "LiteLLM returned an empty transcript." – the recording was too quiet or contained no speech.
• "No document target is available to insert into." – there is no document open to insert into.