Running Daisy fully offline
Daisy can run entirely on your own machine: transcription with bundled whisper.cpp and summaries through a local model server. With both set up, no audio or text leaves your computer.
Set up local transcription
- Go to Settings and open the Providers section.
- Scroll to Local Whisper (offline).
- Under Download a model, choose a size and select Download. The model is saved to your profile, and Daisy stays usable while it downloads.
- Back in Settings → Behavior, set Default transcription provider to
whisper_local.
Available sizes range from tiny through base, small, medium, and large. Smaller models are faster and use less memory; larger ones are more accurate. The .en variants are English-only and slightly faster than their multilingual counterparts.
If you already have a model file on disk, you can skip the download and use …or set a model path manually instead, pointing at the .bin file.
Set up local summaries
- Run a local server such as Ollama or LM Studio, both of which expose an OpenAI-compatible endpoint.
- In Settings → Providers, configure LM Studio or Ollama under Summarization by entering its Base URL (for example,
http://localhost:11434/v1for Ollama). No API key is needed. - In Settings → Behavior, set Default summary provider to that local server.
Good to know
- Offline transcription is semi-live: the transcript arrives in chunks of roughly 15 seconds. Word-by-word live transcription requires Deepgram, which is a cloud provider.
- Local providers are keyless. Daisy checks that the server is reachable rather than verifying any credential.