Nova-3 Streaming
Fast WebSocket transcription with interim results for the lowest-friction everyday dictation flow.
Preview
Native dictation for Mac and iOS with downloaded WhisperKit batch models, sherpa-onnx local streaming, local LLM post-processing, and cloud engines like Deepgram, AssemblyAI, OpenAI, ElevenLabs, Soniox, and Modulate when you want maximum accuracy.
Stay offline with local models, stream for instant feedback, or use a cloud model for specialised accuracy. JustSpeakToIt keeps each path explicit so you always know where audio and text are processed.
Fast WebSocket transcription with interim results for the lowest-friction everyday dictation flow.
High-accuracy multilingual streaming with formatted turn finalisation when you want cleaner long-form output.
Streaming transcription with built-in noise reduction using your existing OpenAI API key.
Real-time Scribe transcription for teams already using ElevenLabs speech workflows.
Low-latency multilingual WebSocket STT for fast feedback across more languages.
Multilingual live transcription with diarization and signal-detection oriented provider support.
Private local recognition for instant, free dictation when cloud accuracy is not required.
Version 2.0 separates local batch, local streaming, and local post-processing so you can pick the right offline tool without guessing which runtime is active.
Download WhisperKit/Core ML models, record normally, then transcribe locally after capture for private drafts, notes, and sensitive dictation.
Use sherpa-onnx streaming models for local live text. The settings UI now shows catalogue/source sections, model sizes, and runtime state consistently.
Download local LLMs for cleanup and formatting. Small models can be less obedient, so the app makes the prompt path visible and keeps built-in rules honest.
A native app that feels immediate. No Electron, no bloat — just fast streaming transcripts, safe key storage, and polished text insertion.
Double-tap Fn, hold-to-record, or use a custom shortcut. Watch live text arrive as you speak, then insert the final transcript where your cursor is.
Keep sensitive recordings offline with Apple Speech, WhisperKit batch models, sherpa-onnx streaming, and local LLM cleanup.
Transcripts are automatically typed into any app using accessibility APIs. Insert at cursor or replace selection—works with any text field.
Optional Live Polish and post-processing using downloaded local LLMs or GPT-4o, Claude, Gemini, and Mistral--removes filler words, fixes punctuation, and polishes your prose.
Have voice-powered conversations with AI. Speak your questions and hear responses read back—complete with conversation history.
Full voice-to-voice conversation loop: record, send, listen to the response, and auto-resume recording. No tapping required.
Hear responses with natural voices from OpenAI, ElevenLabs, Deepgram, Azure, or Apple's built-in system voices.
Teach the app your jargon, names, and acronyms. Your custom dictionary improves accuracy across all providers.
Your transcription history syncs seamlessly between Mac and iOS via iCloud. Pick up right where you left off on any device.
Double-tap Fn or click the menu bar icon. A subtle HUD appears with your active live model.
Just talk naturally. Streaming providers return interim text while you are still speaking.
Stop recording and your polished text appears instantly in any app, right at your cursor.
Download local models for offline work, then use your existing API keys from supported live transcription providers when cloud accuracy, diarisation, or specialist languages matter. Pay only for what you use--typically $0.36 per hour or less with low-cost streaming providers. No subscriptions, no hidden fees, no data selling.
Built with SwiftUI for the best experience on every Apple device.
Yes! When using on-device transcription with Apple Speech, your audio never leaves your device. When using downloaded local models or Apple Speech, audio stays on your device. When using cloud providers like Deepgram, AssemblyAI, Soniox, ElevenLabs, OpenAI, or Modulate, audio is sent directly to their APIs using your own API keys—we never see or store your audio.
The app is free to download. On-device transcription and downloaded local models are free to run after installation. Cloud transcription with your own API keys typically costs around $0.006/minute (about $0.36/hour) with low-cost streaming providers such as Deepgram. You pay the provider directly—we don't add any markup.
Local modes include Apple Speech, downloaded WhisperKit batch models, sherpa-onnx local streaming, and downloaded local LLM cleanup. They prioritise privacy and offline use. Cloud providers like Deepgram, AssemblyAI, Soniox, ElevenLabs, OpenAI, and Modulate offer higher accuracy, lower-latency streaming, broader language support, and provider-specific features like diarisation or formatted turns.
Larger local LLMs tend to follow cleanup instructions better. Very small downloaded models are useful for private, lightweight cleanup, but they may ignore strict formatting requests. The app shows the prompt path so you can tell whether the issue is model capability or configuration.
On Mac, you can capture system audio from meeting apps like Zoom or Teams. On iOS, the app captures audio from your device's microphone—place your phone near the speaker for best results, or use the hands-free conversation mode to interact with AI directly.
Open the app's Settings and enter your API keys in the Providers section. Keys are stored securely in your Mac's Keychain. We never see or transmit your keys—they go directly to the provider's API.
OpenClaw is the built-in AI chat feature. You can have voice-powered conversations with AI assistants— speak your questions, get spoken responses, and use hands-free mode for a completely voice-driven experience. Conversations are saved locally and can sync via iCloud.
Download JustSpeakToIt 2.0 and pick the transcription path that fits the moment: offline local, instant live, or cloud accuracy.
Have a feature request, bug report, or want to contribute? Open an issue on GitHub — we'd love to hear from you!
