SobottaAI
Your voice, your device, your text.
Press a hotkey. Speak. Your words appear anywhere — transcribed locally, enhanced by AI, pasted instantly.
See it in action
Ready to record
How it works
From voice to text in seconds
Record
Hold your hotkey and speak naturally
Transcribe
AI converts speech to text on your device
Enhance
Optional AI polish: emails, code prompts, summaries
Paste
Text appears in any app automatically
Features
Everything you need to dictate like a pro
Privacy First
All transcription runs locally on your device. No audio ever leaves your machine. Your voice data is yours alone.
9 STT Engines
Choose from 5 local Whisper models, 2 NVIDIA Parakeet models, or 2 cloud APIs. From 78MB tiny to 1.6GB flagship.
AI Post-Processing
Transform dictation into professional emails, code prompts, summaries. Use OpenAI, Anthropic, Groq, or fully-local Ollama.
99 Languages
Auto-detect or manually select from 99 supported languages. Built-in support for European, Asian, and Middle Eastern languages.
Paste Anywhere
Transcribed text auto-pastes into any app — your IDE, browser, email client, Slack, anything with a cursor.
GPU Accelerated
Metal acceleration on macOS for near-real-time Whisper inference. Apple CoreML for Parakeet models. Native speed.
Speech-to-Text Engines
9 engines. Your choice.
From a 78MB lightweight model to cloud-powered APIs — pick the right balance of speed, accuracy, and privacy for your workflow.
Whisper Tiny
78 MBFastest, lightweight
99 languages
Whisper Base
148 MBRecommended start
99 languages
Whisper Small
488 MBBalanced accuracy
99 languages
Parakeet TDT 0.6B v2
680 MBNVIDIA, ultra-fast
English
Parakeet TDT 0.6B v3
680 MBNVIDIA, latest
25 European
Whisper Medium
1.5 GBHigh accuracy
99 languages
Whisper Large V3 Turbo
1.6 GBBest quality
99 languages
OpenAI Whisper API
CloudCloud, BYOK
99 languages
Groq Whisper API
CloudUltra-fast cloud
99 languages
AI Functions
Dictation in. Polished output out.
hey john so i was thinking we should probably push the meeting to next week because the design team hasn't finished the mockups yet and it doesn't make sense to review incomplete work what do you think
Hi John, I'd like to suggest we reschedule our review meeting to next week. The design team is still finalizing the mockups, and I believe reviewing completed work will be a more productive use of everyone's time. Please let me know if next week works for you. Best regards
Trust & Transparency
Private by design. Open by default.
100% Local. Zero Compromise.
- Audio never leaves your device
- No cloud accounts required
- Use Ollama for fully offline AI processing
- Bring your own API keys for optional cloud features
MIT Licensed. Fully Transparent.
- Inspect every line of code
- Community-driven development
- No telemetry, no tracking
- Self-host and customize freely
Built with
Powered by best-in-class technology
Tauri v2
10x smaller than Electron
whisper.cpp
OpenAI Whisper, on-device
NVIDIA Parakeet
State-of-the-art STT
Next.js + React
Modern UI
Metal / CoreML
GPU accelerated
SQLite
Local-first storage
Cross-Platform
Native on every desktop
macOS
- Metal GPU acceleration
- Native tray integration
- LaunchAgent autostart
- .dmg installer
Windows
- MSVC native build
- System tray integration
- .msi / .exe installer
- Full feature parity
Linux
- Full ALSA audio support
- .deb and .AppImage packages
- System tray integration
- Wayland & X11 support
Get Started
Start dictating in seconds.
Download SobottaAI — free, open-source, private.
Requires macOS 15+, Windows 10+, or Ubuntu 22.04+