This is a fork of the original opencode-voice. Instead of downloading heavy models and processing audio locally (which consumes CPU/GPU and takes time), this plugin uses Groq's LPU inference engine. Audio is recorded, aggressively compressed to m4a (AAC) on-the-fly, stripped of silence, and transcribed in milliseconds.

Installation

One command through OpenCode:

opencode plugin @loyslow/opencode-voice-groq

Restart OpenCode after installing. First launch will ask for a Groq API key to use their ultra-fast LPU inference engine. The plugin uses ffmpeg automatically.

Optional CLI installer. It runs the same OpenCode plugin install command:

npx @loyslow/opencode-voice-groq install

Do not clone the repo unless you want to develop the plugin.

[!TIP] First launch opens a setup prompt. Enter your free Groq API key, choose your preferred Whisper model, then use ctrl+r to dictate into the prompt.

Requirements

The plugin relies on Groq's API and requires a recording engine:

• uses ffmpeg from your system, or downloads a managed ffmpeg-static fallback
• saves compressed m4a recordings to ~/.cache/opencode-voice-groq/recordings/ temporarily
• uses your provided groqApiKey for transcription

Get your free API key at Groq Console.

Usage

Commands:

• /voice - toggle recording and append transcription
• /voice-submit - toggle recording, append transcription, and submit
• /voice-stop - cancel active recording or transcription
• /voice-settings - open model, hotkeys, microphone, and Groq quotas settings

Default hotkey:

ctrl+r -> start recording
ctrl+r -> stop, transcribe, and append

You can configure a cancel hotkey in /voice-settings.

Models

Available now through Groq API:

| Model | Inference Speed | Notes | | -------------------- | --------------- | ----------------------------- | | Whisper Large v3 | Ultra-fast | highly accurate, multilingual | | Whisper Large v3 Turbo| Insanely fast | slightly faster, great accuracy|

The plugin enforces local RPM tracking (Fail-Fast Quota Protection) to prevent you from recording if you've hit your free API limits, saving you time. Audio is compressed to m4a with silence removal to optimize bandwidth.

Platform Status

| Platform | Status | | -------- | ------ | | Linux | one-command install; recording uses arecord, ffmpeg, or sox | | macOS | one-command install; recording uses ffmpeg AVFoundation | | Windows | one-command install; recording uses DirectShow through a managed cached ffmpeg.exe, with system/bundled ffmpeg fallback |

Architecture

The package follows the public OpenCode TUI plugin shape used by community plugins.

• npm package exports ./tui
• local development can point tui.json at an absolute path
• published install uses opencode plugin @loyslow/opencode-voice-groq
• runtime settings live in OpenCode TUI plugin storage

Files:

• index.js - TUI plugin entrypoint, commands, dialogs, keymap layer, auto-updater
• lib/download.js - utility for fetching the fallback recorder
• lib/engine.js - recorder selection, managed Windows recorder install, and Groq API fetch implementation
• bin/cli.js - install wrapper CLI

Voice input needs native audio recording. The JS plugin manages OpenCode UI, settings, engine downloads, and fast API integration.

Roadmap

• Faster streaming-style transcription via WebSockets
• Further audio compression tuning

Development

Run checks:

npm run prepack
npm pack --dry-run

This MVP has no build step.

Development install from a checkout:

git clone https://github.com/loyslow-dev/opencode-voice-groq.git opencode-voice-groq
cd opencode-voice-groq
opencode plugin "$(pwd)"

Project Status

This is an independent OpenCode plugin. It is not built by the OpenCode team and is not affiliated with OpenCode.

Credits

• OpenCode wordmark SVG adapted from the public OpenCode repository. The voice mark was added for this plugin.
• This is a fork of the original opencode-voice project created by @ihxnnxs.

OpenCode Website | Docs | Discord