Stenographer

Record conversations. Keep the words. Revisit what mattered.

Stenographer 3.0 brings local recording, transcription and source-linked Chat into a cleaner workspace on iPhone and iPad, with Apple Watch recording as a companion.

• Record on iPhone or Apple Watch and keep continuous capture together as one conversation.
• Read transcripts, play original audio, search words and tags, and edit speaker names.
• See live text when processing is available; saved audio remains available when processing is deferred.
• Ask about a conversation or your library. Numbered sources open the captured passages behind an answer.
• Review action targets before opening, playing, sharing or deleting recordings.

LOCAL RECORDING AND TRANSCRIPTION
Audio is recorded locally. Watch audio transfers to the paired iPhone for processing. Small multilingual transcription is included and works offline. Settings also offers downloadable Large multilingual; it needs more memory and processing time. Settings also offers optional Distil English-only transcription. Your multilingual settings are saved for switching back. The app shows the selected and actual model, preparation progress and recovery options. Some model failures can use Small while preserving your Large selection.

LOCAL AI
Apple Intelligence is the default local Chat option on supported, ready devices. You can download and select Qwen3.5-4B for local use. Automatic summaries and tags stay local. Keep the phone app active and idle for pending automatic work to finish. Model availability and speed vary with device, OS, language, downloaded assets and workload.

OPTIONAL CLOUD CHAT
Bring your own OpenAI, Anthropic or OpenRouter API key and choose a provider model. Keys are stored in device Keychain. Explicit consent is required before cloud Chat sends your current question, selected transcript passages and recording dates and identifiers to the selected provider. Audio files and voice profiles are not sent. OpenRouter may route to another provider. Provider data policies and account charges apply. Failed requests do not silently switch providers.

AI, transcripts and speaker attribution can contain mistakes or omissions. Check the original audio and source passages. Sharing uses your chosen destination; deletion requires confirmation.