Say it. Sotto types it, or does it.

Dictation is the core, but Sotto also includes Sotto Assistant (a voice assistant with 77 Mac tools that can see your screen, talk back and use your own MCP servers), a local REST API with a CLI so other machines and scripts can use your Mac's transcription, system-audio recording for meetings, watch-folder automation, Shortcuts actions and text-to-speech.

Hold a key and ask. It has 77 built-in Mac tools covering Calendar, Reminders, Mail, Messages, Notes, Contacts, Maps, Weather, web, image and YouTube search, the browser, Finder, apps and windows, the clipboard, selected text, screenshots, Shortcuts, media and volume, alarms and timers, and policy-gated bash. It can optionally include a screenshot of your screen as context and answer out loud with local, OpenAI or ElevenLabs voices.

Yes. Sotto is an MCP client: add the MCP servers you already run in the Assistant's tool settings and it can call their tools alongside the built-in ones.

Yes. Sotto runs a local REST API (37 endpoints, with an OpenAPI spec and Swagger docs at /docs) protected by per-device bearer tokens. Turn on remote connections to reach it over your Wi-Fi or Tailscale, or use the bundled sotto CLI with --host to talk to another Mac.

No. Local WhisperKit and Parakeet models handle transcription on your Mac, and Ollama, LM Studio and Apple Intelligence run AI locally without a key. You can also route AI through Claude Code, Codex, OpenCode, Grok CLI or Cursor Agent with no API key. Cloud providers are optional and use your own key.

Yes. The default WhisperKit and Parakeet models run entirely on your Mac using Apple's Neural Engine. After a one-time model download, you can dictate and transcribe with no internet connection at all. Cloud models are strictly optional.

With local models, your audio never leaves your Mac — nothing is uploaded, ever. If you choose to connect a cloud model, you bring your own API key and only those recordings are sent to that provider. There's no Sotto server in the middle.

Sotto ships with OpenAI's Whisper models (Tiny through Large V3 Turbo and Distil Large V3) and NVIDIA's Parakeet v2/v3 models, all running locally. For cloud transcription you can connect OpenAI, Groq, xAI Grok, Deepgram Nova-3, AssemblyAI, Cartesia Ink or any OpenAI-compatible endpoint with your own key. You can re-transcribe any saved recording with a different model anytime.

Yes. Sotto can record system audio from meetings, calls and videos, or run continuous mic and system recording. It records on your Mac, so no bot joins the call. You can also import audio and video files or point a watch folder at a directory to auto-transcribe new files with a title and summary.

Yes. $49 once, use it on up to 3 Macs, and all future updates are included. No subscription, no annual renewal, no upsell to a 'pro' tier.

Both are local-first dictation apps for macOS. The big difference is pricing: Sotto is a one-time $49 purchase with lifetime updates on 3 Macs, while SuperWhisper's full feature set is sold as a subscription (with a pricier lifetime tier). Sotto also includes Parakeet models, per-language hotkeys, and AI text-processing functions.

MacWhisper is primarily a file-transcription app: you open it, drop a file in, and get text out. Sotto is dictation-first — a system-wide push-to-talk layer that types into whatever app you're using — and it also imports and transcribes audio files. If you mostly dictate into other apps, Sotto is built for exactly that.

Sotto needs macOS 15 Sequoia or later. It's a native Swift app, and Apple Silicon Macs get the best performance since local models run on the Neural Engine.

90+ languages via Whisper, with Parakeet v3 adding state-of-the-art multilingual accuracy. You can switch languages from the recording bar, assign a dedicated hotkey per language, or let the model auto-detect what you're speaking.