You're in a chat. You'd rather talk than type. Tap a button on your phone, speak, and the words appear wherever your cursor is — any app, any OS, any field. Nymble is your personal typist.
Your phone is already the best microphone you own. Nymble turns it into a voice-powered keyboard for your computer.
Scan a QR code from the Nymble app on your computer. Same WiFi, or across the internet — both work.
Hit the big button in the mobile app. Talk naturally. Your speech streams to a transcription engine in real time.
The transcribed text is typed into whatever's focused on your PC — character by character, like a real keyboard. Because it is one.
Most dictation tools paste text or use virtual keyboards — and apps handle those inconsistently. Some block paste, some ignore virtual input, and accessibility permissions vary by OS. It's always a compromise.
Nymble's relay types every word through a real HID keyboard device. To your computer, it's indistinguishable from someone physically pressing keys. If an app accepts typing, it accepts Nymble. No exceptions.
Chat apps, terminals, IDEs, browsers, password managers — if you can click and type, Nymble works there.
No accessibility hacks, no admin rights, no IT tickets. It's a keyboard — your OS already trusts it.
Corporate PCs where you can't install software? Plug in the HID device. Done. No drivers needed.
Windows, macOS, Linux, ChromeOS — even BIOS screens. If it takes a USB keyboard, it takes Nymble.
Use your phone for the fastest path, or a dedicated Pi for an always-on offline setup. Both send keystrokes through the same relay.
A cross-platform app (iOS & Android) with one big button. Tap, speak, and your words stream through cloud-based transcription to your PC in real time. Pair once, use from anywhere.
A Raspberry Pi with a dedicated microphone running OpenAI Whisper locally. Your voice never leaves the device. Ideal for always-on desk setups and privacy-first environments.
Nymble's speech-to-text layer is provider-agnostic. Pick the option that matches your priorities — speed, privacy, or both.
Real-time transcription via Deepgram, AssemblyAI, or other streaming providers. Lowest latency, text appears as you speak. Best for the mobile app workflow.
Run Whisper on your own server via any OpenAI-compatible API (LocalAI, Ollama, etc.). Your audio never leaves your network. Point the config at your local endpoint and go.
Whisper running directly on the Raspberry Pi. No network, no cloud, no data leaving the room. Air-gapped transcription for the most sensitive environments.
QR-based local pairing for same-network setups. Discovery server for cross-network — your phone on 5G can dictate to your PC on home WiFi.
With streaming STT providers, text appears on screen as you speak. No waiting for the recording to finish — it's live dictation.
Each piece is independent. Mobile app, relay tray, HID device, Pi, discovery server — use what you need, swap what you don't.
Transcription happens on your phone (cloud) or your Pi (local) — never on your PC. No GPU hogging, no background processes eating RAM.
Server-side components ship with Docker and Compose files. One command to deploy the discovery server or the Pi transcription engine.
Every line of code, every design decision — open and auditable. Contribute, fork, or build your own configuration.
| Product | Real Keyboard | Any App | Any OS | Offline Option | Open Source | Mobile Input |
|---|---|---|---|---|---|---|
| Nymble | ✓ HID | ✓ | ✓ | ✓ | ✓ | ✓ |
| Superwhisper | Paste | ✓ | Mac only | ✓ | ✗ | ✗ |
| Wispr Flow | Paste | ✓ | Mac only | Partial | ✗ | ✗ |
| Windows Voice Typing | Virtual KB | Limited | Windows | Partial | ✗ | ✗ |
| OpenWhispr | Paste | ✓ | Win/Mac | ✓ | ✓ | ✗ |
| Dragon / Nuance | API | ✓ | Windows | ✓ | ✗ | ✗ |
Slack, Discord, Telegram, Mattermost — you're in conversations all day. Hit the button, speak your message, and it's typed out. No more thumb fatigue.
Dictate prompts to Cursor, Copilot, or Claude at 200+ WPM. Explain your intent at the speed of thought, not the speed of your keyboard.
RSI, carpal tunnel, motor disabilities — Nymble turns any computer into a voice-first machine without installing anything on the host.
Journalists, lawyers, healthcare — anyone who needs air-gapped transcription. Run it all on local hardware. Your voice never leaves the room.
Each component lives in its own repository under NymbleType on GitHub. Clone what you need, contribute where you can.
Cross-platform mobile app (Flutter). Big-button voice capture, streaming STT, device pairing — your phone as a wireless keyboard mic.
Flutter · iOS · AndroidCross-platform system tray app that receives transcribed text and routes it to the HID device for typing. The bridge between voice and keys.
Desktop · Win · Mac · LinuxRP2040 firmware that emulates a USB HID keyboard. Receives text from the relay and types it out — character by character, at hardware level.
C · RP2040 · USB HIDProvider-agnostic STT abstraction layer. Deepgram streaming, AssemblyAI, OpenAI Whisper batch, and self-hosted Whisper-compatible endpoints.
STT · Provider InterfaceOffline Whisper transcription engine for Raspberry Pi. Mic capture, VAD, real-time speech-to-text — fully air-gapped. Docker-ready.
Python · Raspberry Pi · DockerLightweight WebSocket relay broker for cross-network device pairing. Phone on 5G, PC on WiFi — the discovery server bridges them.
Node.js · WebSocket · DockerCore repos live. Mobile app scaffolded with streaming STT and pairing. Relay tray and HID firmware functional. Discovery server running with Docker support.
Full mobile → relay → HID pipeline working. Mic capture, real-time Deepgram streaming, QR pairing, and cross-network discovery server integration. First real dictation sessions.
Multi-language support, latency optimization, connection resilience, relay tray UI refinements, and improved VAD. Self-hosted Whisper provider testing.
Targeted input destinations — route text to specific apps from the mobile UI. Multiple paired devices. Pi hardware kit with build guide. Community contributions and plugin system.