Voice.nvim

A Neovim plugin for voice recording and transcription using Whisper models.

Prerequisites

Before installing the plugin, ensure you have the following dependencies:

• Neovim >= 0.8.0
• sox (for audio recording)
• ffmpeg (for audio processing)
• cmake and make (for building whisper.cpp)
• git (for installation)

Installing Prerequisites

macOS

brew install sox ffmpeg cmake make git

Linux (Ubuntu/Debian)

sudo apt-get update
sudo apt-get install sox ffmpeg cmake make git

Installation

Using lazy.nvim

{
    "ronjansen/voice.nvim",
    dependencies = {
        "nvim-lua/plenary.nvim",
    },
    config = function()
        require("voice").setup({
            -- your configuration here
        })
    end
}

Using vim-plug

Plug 'nvim-lua/plenary.nvim'
Plug 'ronjansen/voice.nvim'

Then add to your init.lua:

require("voice").setup({
    -- your configuration here
})

Configuration

Here's the default configuration:

require('voice').setup({
    model = 'small',                -- Whisper model size
    backend = 'whisper.cpp',        -- Backend ('whisper.cpp' or 'openai')
    keymaps = {
        toggle_recording = '<leader>r', -- Keybinding for recording
    },
    max_recording_seconds = 60,     -- Max recording duration
    openai = {
        api_key = nil,              -- Your OpenAI API key if using OpenAI backend
    },
    ui = {
        show_recording_status = true,   -- Show recording status in statusline
        floating_window = true,         -- Show transcription progress in floating window
    }
})

Usage

1. Start recording by pressing <leader>r (or your configured keybinding)
2. Speak clearly into your microphone (English is recommended)
3. Press <leader>r again to stop recording
4. Wait for transcription to complete
5. The transcribed text will appear at your cursor position

Troubleshooting

Common Issues

License

MIT License - See LICENSE file for details