Juno is a native Mac voice layer for people who want speech to become finished work, not just a raw transcript. Press a hotkey, talk naturally, watch the live transcript, and let Juno commit the final text into the app you were already using.
It is designed to be local, private, and free to run from source. The runtime handles live preview, final transcription, writing cleanup, transformations, actions, dictionary and memory, app context, privacy gates, and native insertion without requiring a hosted transcription account.
Juno next to Apple Dictation, Wispr Flow, and Superwhisper, across the loop that matters: live transcript, cleaned writing, Mac actions, rewrites, and price.
| Difference | Juno | Apple Dictation | Wispr Flow | Superwhisper |
|---|---|---|---|---|
| Live HUD transcript | Yes. Live HUD transcript while dictating. | No. Text appears in the current field. | No. Voice-to-text across apps. | No. Voice-to-text across apps. |
| Final cleaned writing | Yes. Pauses, restarts, and corrections become finished text. | Basic dictation commands and punctuation. | Yes. Positioned around polished voice writing. | Yes. Custom AI modes and formatting. |
| Mac-native actions | Yes. Notes, Reminders, and alarms from voice. | No. Dictation enters text. | No. Not the product center. | No. Not the product center. |
| Selected text rewrites | Yes. Rewrite selected or recent text in place. | No. Built-in dictation is not a rewrite layer. | Yes. Command mode is in paid Pro. | Yes. Custom prompts and AI modes. |
| Price model | Free forever. No paid tier, no usage cap. | Included with macOS. | Free Basic has word limits. Pro is paid. | Free tier. Pro and Enterprise are paid. |
| Open source | Yes | No | No | No |
| Best reason to pick it | You want private Mac work, not another SaaS meter. | You only need built-in speech-to-text. | You want a polished cross-platform voice-writing service. | You want a mature dictation app with meeting/file workflows. |
The live version of this comparison is at usejuno.co/#compare.
- Writes into the active Mac app from a hotkey.
- Shows live words while you are still speaking.
- Produces a cleaner final transcript when you stop.
- Turns rough speech into paragraphs, bullets, replies, notes, and structured writing.
- Rewrites selected or recent text with spoken commands.
- Creates notes, reminders, and alarms from natural language.
- Learns local vocabulary, names, snippets, replacements, and corrections.
- Uses app context so chat, email, notes, documents, code, and terminal surfaces are handled differently.
- Applies privacy gates for sensitive fields, capture, history, learning, and insertion.
Juno can route speech into actions instead of only inserting text.
Hey Juno, note that the design review moved to Thursday.
Hey Juno, remind me tomorrow at 9 to send the agenda.
Hey Juno, set an alarm for 6:30.
One utterance can become clean writing plus follow-up work. Juno can split compound requests, resolve dates, and send the result to the local action sinks available on your Mac.
Juno is local-first by design.
- Microphone capture starts only when you trigger dictation or enable an explicit listening mode.
- Runtime data is stored locally on your machine.
- Dictionary and memory are local product features, not a cloud profile.
- Secure or sensitive surfaces can suppress context, learning, history, audio retention, and paste.
- The source runtime does not require an account or per-minute cloud billing.
- macOS 15 or newer for the native shell.
- Apple Silicon Mac for the full local model path.
- Python 3.10 or newer.
- Microphone permission for dictation.
- Accessibility permission for native insertion into other apps.
Linux can run selected runtime checks and portable Python paths, but the shipping desktop product is macOS-first.
Bootstrap the Python environment:
./scripts/bootstrap.sh
source .venv/bin/activateCheck the environment:
./scripts/doctor.sh --ciInstall the default local model assets:
./scripts/bootstrap_full.shStart the standalone workbench:
./scripts/run_workbench.shRun the local Mac voice stack:
./scripts/run_live.shInstall the macOS app locally:
./scripts/install_macos.sh --install-to-appsPackage the macOS app:
./scripts/package_macos.sh
Juno follows one product path:
audio -> speech state -> live preview -> final transcript -> writer -> actions -> native insertion
The Mac shell owns hotkeys, permissions, window state, secure-field policy, and insertion. The local runtime owns speech processing, preview, final transcription, writing, actions, dictionary and memory, app context, history, and health reporting.
Run the public smoke check before publishing changes:
./scripts/smoke_test.shUseful entry points:
scripts/bootstrap.shsets up the Python environment.scripts/doctor.shchecks required, optional, and platform-specific setup.scripts/run_workbench.shstarts the local workbench server.scripts/run_live.shstarts the local Mac voice stack.scripts/install_macos.shinstalls the app locally.scripts/package_macos.shpackages the app.
Core folders:
juno_v2/contains the voice runtime, workbench, memory, context, preview, final transcription, writer, and health tools.juno_core_v3/contains broker contracts, actions, policy, model registry, and compatibility layers.shells/macos/contains the native Mac shell.seed_data/contains local vocabulary and personalization seed data.config/contains example local configuration.
Build an OTA release for Sparkle:
./scripts/generate_juno_sparkle_keys.sh
./scripts/build_juno_ota_release.sh --version 0.2.1 --build-number 2 --ota-feed-url https://updates.example.com/juno/appcast.xml --ota-public-ed-key "$JUNO_OTA_PUBLIC_ED_KEY" --download-url-prefix https://updates.example.com/juno/The OTA release script builds dist/Juno.app, creates a signed Sparkle archive
under dist/ota/, and regenerates dist/ota/appcast.xml. Use
--allow-insecure-ota-feed only for local http or file appcast testing.
For the full setup checklist, see Juno OTA Updates.
MIT. See LICENSE.