Does it label speakers or add timestamps?

No. It produces a single transcript without speaker labels or timestamps.

Is the recording uploaded?

Not with a local model. With cloud transcription the file is sent directly to your chosen provider.

What if a file is very long?

WhisprDesk does not enforce a size limit for local transcription, but very long files have not been tested. Try a shorter file first.

Transcribing interviews on Mac, with local transcription

An interview recording is often something you would rather not upload anywhere. With a local model, WhisprDesk transcribes the file on your Mac: drop the recording into Upload, and the transcript appears as a note.

From recording to transcript

Open Upload, drop in the file, and click Transcribe. Accepted formats are MP3, WAV, M4A, WebM, OGG, FLAC, AAC, MP4, WMA and Opus. The result is saved in Notes, where you can copy it or save it as a Markdown file.

WhisprDesk reads the recording where it is and does not copy or keep it. Only the transcript is saved. Keep your own backup of the original audio.

Choosing local for sensitive material

With a downloaded local model, the file is processed on your Mac and is not uploaded. With cloud transcription, the file goes directly to the provider whose key you entered and is limited to 25 MB. WhisprDesk does not receive either one. The choice is yours, and you can change it at any time in Settings > Transcription.

Local models need a one-time download: NVIDIA Parakeet is about 680 MB, and Whisper models range from about 75 MB to about 3 GB. You can transcribe without an internet connection once a model is downloaded.

Names and terms

Add the names of people, organizations and places to the Dictionary to help Whisper and cloud models recognize them. The Dictionary does not apply to NVIDIA Parakeet.

Check every quote

A transcript is a working draft. Listen back to anything you intend to quote. WhisprDesk produces one block of text and does not label who is speaking, so you will add speaker names yourself.

Frequently asked questions

Does it label speakers or add timestamps?: No. It produces a single transcript without speaker labels or timestamps.
Is the recording uploaded?: Not with a local model. With cloud transcription the file is sent directly to your chosen provider.
What if a file is very long?: WhisprDesk does not enforce a size limit for local transcription, but very long files have not been tested. Try a shorter file first.