sybl 0.1.0

Open-source BYOK voice dictation — speak anywhere, transcribe fast, type it in.

sybl

Open-source, bring-your-own-key voice dictation. Put your cursor anywhere, hold a global shortcut, speak, and sybl transcribes it fast — then types it in for you.

The open-source alternative to closed dictation tools like Wispr Flow. No subscription, no sybl-hosted backend: your audio goes straight to the STT provider you choose.

Features

• Global push-to-talk — hold a system-wide hotkey, speak, release to finish
• BYOK STT — Groq Whisper (batch) and Deepgram (streaming) today; pluggable providers
• Types where you were focused — clipboard-paste injection on Windows
• Daemon + TUI — background service with a Textual dashboard for logs, history, and settings
• Listening pill — small overlay near your cursor while dictating (Windows)
• Custom vocabulary — STT hints for names and jargon via sybl config vocab
• Voice commands — new line, period, comma in final transcripts
• Local-first — no telemetry; API keys in the OS keyring

Install

pipx install sybl

Or with uv:

uv tool install sybl

Requires Python 3.12+. Windows is the primary supported platform for the full core loop (hotkeys, injection, indicator).

Quick start

sybl doctor
sybl config init
sybl config set-key groq
sybl start

In another terminal:

sybl tui

Hold Ctrl+Alt+Space (default), speak, release — the transcript is pasted into whatever app had focus. Press Esc while holding to cancel.

See Getting started for the full walkthrough.

Documentation

Doc	Description
Getting started	Install, first run, daemon + TUI
Providers	Groq & Deepgram BYOK setup
Configuration	config.toml reference
Permissions	Microphone and injection notes
Development	Local dev, phases, integration tests
Roadmap	What's built and what's next
Daemon architecture	IPC and process model

Contributing

Contributions welcome! See CONTRIBUTING.md and AGENTS.md for architecture and conventions.

License

MIT © Rikhil Nellimarla