opencode-voice

Voice control for OpenCode

Everything runs on your machine — no cloud services, no API keys, no audio leaves your device.

Microphone → whisper.cpp → text → OpenCode

---

Install

cargo install opencode-voice

Requires Rust 1.70+, cmake, and a C compiler. See Building from source for details.

Setup

Download a transcription model (~150 MB):

opencode-voice setup

Run

Start OpenCode with its HTTP server, then start opencode-voice in a separate terminal:

opencode --port 4096
opencode-voice --port 4096

Hold space to record, release to transcribe and send. Press q to quit.

---

Options

Flag	Default	Description
--port <n>	(required)	OpenCode server port
--model <size>	base.en	Whisper model (see below)
--device <name>	system default	Audio input device
--key <name>	space	Toggle key for recording
--hotkey <name>	right_option	Global hotkey (system-wide)
--no-global	—	Disable global hotkey
--no-push-to-talk	—	Toggle mode instead of hold-to-talk
--no-auto-submit	—	Don't auto-submit; leave transcript in prompt for review
--no-handle-prompts	—	Don't handle OpenCode permission/question prompts via voice

Models

English-only models are fine-tuned on English and generally more accurate. Multilingual models support 99 languages and may help with some accents, but the biggest improvement for accent robustness comes from using a larger model size.

Model	Type	Size
tiny.en	English-only	~75 MB
base.en	English-only (default)	~142 MB
small.en	English-only	~466 MB
tiny	Multilingual	~75 MB
base	Multilingual	~142 MB
small	Multilingual	~466 MB

Environment variables: OPENCODE_VOICE_PORT, OPENCODE_VOICE_MODEL, OPENCODE_SERVER_PASSWORD.

Other commands

opencode-voice setup [--model <MODEL>]   Download whisper model
opencode-voice devices                   List audio input devices
opencode-voice keys                      List key names for --key / --hotkey

Troubleshooting

Problem	Fix
Can't connect to OpenCode	Make sure OpenCode is running with --port
No speech detected	Speak closer to mic, try --model small.en, check opencode-voice devices
Whisper model not found	Run opencode-voice setup
Global hotkeys don't work (macOS)	Grant Accessibility permission in System Settings, or use --no-global
Auth errors (401)	Set OPENCODE_SERVER_PASSWORD env var
Build failures	See Building from source

Platform support

Platform	Status
macOS	Fully supported (requires Accessibility permission for global hotkeys)
Linux	Supported (ALSA or PulseAudio)
Windows	Not supported

License

Apache 2.0 — see LICENSE.