Capture anywhere
Web, macOS, and Windows record microphone audio, optional system audio, or imported files. Desktop is the uninterrupted-recording reference client.
Record microphone and device audio, keep working offline, and turn the day into searchable transcripts and useful memories. Audio is removed after local transcription.
Independently decodable offline blocks with idempotent upload.
Transcript persistence + unlink receipt before client release.
Web, macOS, and Windows record microphone audio, optional system audio, or imported files. Desktop is the uninterrupted-recording reference client.
CPU-only speech recognition, VAD, diarization, and multilingual embeddings run without paid speech services.
Search transcripts and memories instantly with FTS5 and semantic vectors, or ask a question and get an answer written over cited local context.
The client retains each chunk until transcript data is durably committed and the server confirms its temporary file was deleted. NeoRecall keeps transcripts, memory data, and an explicit 5–10 second speaker-only preview per recognized voice—not the original recordings.
Meetings and important conversations become episodic memories. Facts, people, tasks, promises, events, and relationships remain atomic and easy to act on.
Original-language transcripts stay intact while consolidated memory text is normalized to English for consistent recall across German and English days.
First-run consent guidance, persistent indicators, and immediate pause controls keep capture explicit. NeoRecall does not include a covert mode.
Setup downloads pinned local models, verifies native inference and sqlite-vec, then installs a user service. Docker Compose is available as an alternative.
npm i -g neorecall
neorecall install
neorecall setup
neorecall start