Skip to content

Reference

Privacy & offline

Understand what stays local, what uses a network, and how to work offline.

On this page

OpenWhisper supports local transcription, cloud transcription, and a paired remote engine. Network behavior depends on the speech engine and any optional text provider you choose.

Understand each connection#

FeatureWhat uses a network
Local speech modelInitial model and runtime downloads. Installed models load from the local cache without network metadata checks.
Cloud transcriptionAudio is sent to the selected cloud provider. An API key and internet connection are required.
Remote engineAudio is sent to your authorized host computer for transcription.
AI cleanupTranscript text goes to the selected text endpoint. A local endpoint can keep this processing on your computer.
Meeting intelligenceThe configured endpoint or agent processes meeting content and may use its provider's network services.
Host record storageRecords transfer to the paired host when host storage or Both is enabled.

Selecting a local speech engine does not make cloud AI cleanup local. Check both the speech engine and text provider for a fully local workflow.

Prepare for offline use#

  1. Choose a local speech engine.
  2. Download its model and every required runtime in Settings → Downloads while online.
  3. Turn off cloud cleanup and cloud meeting intelligence, or use a locally available text endpoint.
  4. Make a short test recording after disconnecting from the internet.

Installed models load from the local cache without checking online metadata. A remote engine still needs a reachable host even when the internet is unavailable.

Control missing-model downloads#

In Settings → Downloads → When a model is missing from this computer, choose whether to ask, allow, or block downloads.

For source launches, setting HF_HUB_OFFLINE=1 before launch blocks Hugging Face model downloads. It does not disable cloud transcription or network text providers.

Know where credentials and records live#

API keys added in Settings use your operating system's credential store. History, transcripts, and recordings are kept by the app; remote-engine storage can additionally keep records on the host.

Review Where records are kept when configuring a remote engine. See Remote engines for the choices.

Exports may contain transcript content and recording references. Share them according to the content they contain; a JSON export does not include the referenced audio files.

Review downloaded components#

Downloads shows component publishers, source hosts, versions, licenses, and integrity checks. Integrity checks confirm that a download matches the expected artifact; they do not guarantee model or runtime security.

Keep the app and its runtimes updated and review important transcripts for errors. Use Choosing a model to understand platform support and experimental options.

Checked against OpenWhisper 2.6.14.App documentation
Still stuck?

Get help from the community, or tell us what went wrong.