Stenographer
privacy policy.

Recording and local processing

Stenographer records audio on iPhone and Apple Watch. Audio, transcripts, summaries, topic labels and remembered voice information are stored in the app's local storage. Watch recordings transfer to the paired iPhone for processing and may wait for connectivity.

Whisper transcribes audio on iPhone. Translation to English is an explicit setting. Apple Intelligence provides local AI on supported, available devices; an optional Qwen model can be downloaded for local use. Automatic summaries and topic labels use local AI. Availability depends on your device, operating system, language and model readiness.

Naming a speaker can create local voice information used to suggest that name in later conversations. Matching is approximate. You can review, correct, unlink or forget remembered voices. These voice profiles are not sent to AI providers.

Downloading an optional local model uses the internet. Download hosts receive ordinary network-request information, such as your IP address, but these downloads do not include your recordings. After installation, local inference does not send recording content to a cloud AI provider. Apple controls the availability and management of its system models.

Optional cloud Chat in version 3.0

You can choose OpenAI, Anthropic or OpenRouter using your own API account and key. Before activating cloud Chat, you choose a provider and model and consent to sending your question, relevant chat context, selected transcript passages and recording details, such as dates and identifiers, to that provider.

Home Chat can use context from your library. Conversation Chat is scoped to the selected conversation. A requested Chat summary may send that conversation's transcript in parts. Selecting a recording through Chat may send candidate dates, identifiers and short excerpts. Text can contain personal or sensitive information from your conversations.

Audio files and speaker profiles are not sent to these AI providers. Requests go directly from the app to the selected provider. The provider can associate requests with the API account you use. Provider charges, processing locations, retention policies and account settings apply. OpenRouter may route requests to another model provider.

Some providers retain request content or related logs beyond the request, and some downstream providers may use content to improve models. Stenographer does not promise zero retention or a single training policy for every provider, account or route. Review the applicable policies before consenting:

• OpenAI API data controls
• Anthropic API retention
• OpenRouter privacy policy and the policies of its selected downstream provider

Model discovery sends your API key to the selected provider to request its model catalog; it does not send recording content. Keys use nonsynchronizing device Keychain storage and are not included in transcript exports. Removing a key or consent prevents further use through that configuration. A failed cloud request does not silently switch providers. Selecting local AI keeps subsequent local requests on the local route.

Sharing, backups and deletion

Sharing uses the operating system's share sheet and the destination you choose. Exported JSON contains transcript and recording metadata; it is not an audio backup. Device and operating-system backup settings may copy app data separately from AI processing.

Use the app's conversation-deletion and remembered-voice controls to manage local data. Conversation deletion requires confirmation. Forgetting a remembered voice preserves the original audio and transcript words. You can change microphone access in system Settings.

Deleting local data, removing a key or withdrawing consent does not erase copies already sent to providers, shared destinations or backups. Manage those copies through the relevant provider, destination or backup controls.

Questions and support

Contact hello@sleepingbear.dev about Stenographer privacy or support. If you send us a support message, we receive the information and attachments you choose to include. Only include recording content when you intend to share it for support.