Crash-safe
Every session is written as raw PCM WAV with the header flushed every few seconds, so a crash still leaves usable audio behind.
Stenobar
Catch what you hear. Keep what you say.
Stenobar captures system audio and your mic at the same time, then transcribes, summarizes, and tags every recording into a library you can search by what was said.
Coming soon
Native Swift & SwiftUI macOS 26+ Apple Silicon only
Record now. Transcribe and summarize only if you want.
Stenobar is a recorder first. Everything else is optional. Transcription and AI summaries point at on-device engines or your own cloud keys, or you never turn them on.
Record
Capture system audio and your mic as two separate tracks, straight from a menu-bar item. Grab everything the system plays, or just the apps you pick: Zoom, Safari, whatever you choose.
Transcribe Optional
Turn any recording into searchable text. Run it entirely on-device with Apple SpeechAnalyzer or WhisperKit, or point it at a cloud engine like AssemblyAI, Deepgram, OpenAI, Groq, or Google. Completely optional: with nothing configured, your recordings are still yours to play, export, and keep.
Summarize & tag Optional
Hand a transcript to an LLM for a summary and suggested tags, whether that's Anthropic, OpenAI, OpenRouter, or a local Ollama model. It's optional too, and separate from transcription, so you can mix providers or skip it altogether.
Dictate
Hold a hotkey, speak, release. A floating HUD shows live text that is transcribed on-device and pasted straight into whatever app you are using.
Thoughts
A second hotkey for fleeting ideas: speak, and Stenobar routes the result to Reminders, Notes, Todoist, a Shortcut, or a local inbox.
Search everything
A three-pane library with full-text search across titles, tags, summaries, and transcript bodies. Filter by date, sort, star, and group by project.
The details that keep it solid.
Projects
Group recordings into project folders, each with its own transcription overrides. Drag between projects and the files follow on disk.
Your format
16- or 24-bit, 44.1 or 48 kHz, mono or stereo. Export audio to M4A or FLAC and transcripts to TXT, Markdown, SRT, or VTT.
Combined track
Optionally stitch system and mic into one mixdown (mono, split stereo, or mixed), and re-stitch any time you change your mind.
Import audio
Drop in existing WAV, MP3, M4A, AIFF, FLAC, or CAF files to transcribe, tag, summarize, and search alongside your recordings.
Tidy on its own
Optional retention trashes old, unstarred recordings after a set number of days. Starred recordings always stay, and deletes go to the Trash.
Bring your own engines and keys.
Stenobar doesn't tie you to one vendor. Run everything on-device for full privacy, or drop in your own keys for the cloud services you already use — per recording, and only when you choose. Transcription and AI are picked separately, so you can mix and match freely.
Cloud
Cloud transcription
- AssemblyAI
- Deepgram
- OpenAI
- Groq
- Google Cloud
On-device
On-device transcription
- Apple SpeechAnalyzer
- WhisperKit
- Parakeet
Whisper tiny.en → large-v3 · Parakeet TDT 0.6B (v2 / v3)
Cloud
Cloud AI — summaries, tags & thoughts
- OpenAI
- Anthropic
- OpenRouter
- OpenAI-compatible
On-device
On-device AI
- Apple Intelligence
- Ollama
Any local Ollama model, on your machine
Your recordings stay on your Mac.
Local transcription and Thoughts routing never leave the device. Cloud providers are strictly opt-in and only ever touch the single recording you ask them to process. Your API keys live in the macOS Keychain, nowhere else. No analytics. No tracking. No account.
- On-device by default
- Keys in Keychain
- Opt-in cloud, per recording
- Zero telemetry