Vocord

Cross-platform voice message transcription for Vencord. 100% local, privacy-first.

Quick Start

macOS / Linux:

curl -sSL https://raw.githubusercontent.com/jolehuit/vocord/main/install.sh | bash

Windows (PowerShell):

irm https://raw.githubusercontent.com/jolehuit/vocord/main/install.ps1 | iex

That's it. The installer handles everything automatically:

• Finds your Vencord source tree (or clones it if you don't have one)
• macOS ARM: lets you choose between mlx-whisper (GPU, recommended) and whisper-onnx (CPU)
• Linux / macOS Intel / Windows: installs Rust if needed, builds the transcribe-cli binary, downloads the Whisper model
• Cleans up any existing Vocord installation before reinstalling
• Copies the plugin into Vencord and rebuilds (verifies Vocord is in the build output)
• Auto-configures Vesktop if detected (closes it, sets the custom Vencord build path, restarts it)
• Auto-injects into Discord Desktop if Vesktop is not found

Works with Vesktop, Discord Desktop + Vencord, and Equicord on macOS, Linux, and Windows.

Restart Discord / Vesktop, enable Vocord in Settings > Vencord > Plugins, and you're good to go.

Tip: If the installer can't find your Vencord source, you can specify it:

VENCORD_DIR=~/path/to/Vencord curl -sSL https://raw.githubusercontent.com/jolehuit/vocord/main/install.sh | bash

On Windows:

$env:VENCORD_DIR="C:\path\to\Vencord"; irm https://raw.githubusercontent.com/jolehuit/vocord/main/install.ps1 | iex

How It Works

Platform	Backend	Model
macOS Apple Silicon	mlx-whisper	whisper-large-v3-turbo — GPU via Apple MLX
Linux / Windows / macOS Intel	transcribe-rs (Whisper GGML)	ggml-large-v3-turbo — CPU/GPU

The installer lets you choose your backend on macOS Apple Silicon. On other platforms, transcribe-rs is used automatically.

Auto-detects language — no configuration needed.

Features

• One-click transcription of voice messages
• 100% local — no data sent to external servers
• Auto language detection
• Copy transcription to clipboard
• Clean reinstall — re-running the installer removes the old version automatically

Settings

Setting	Description	Default
Show Toast	Notification on completion	true

Everything (backend, model, ffmpeg) is auto-detected. No manual configuration needed.

Manual Install

If you prefer not to use the one-liner:

1. Clone Vencord (if not already):

   git clone https://github.com/Vendicated/Vencord && cd Vencord && pnpm install

2. Clone and copy Vocord:

   git clone https://github.com/jolehuit/vocord.git /tmp/vocord
   mkdir -p src/userplugins/vocord
   cp /tmp/vocord/index.tsx /tmp/vocord/native.ts src/userplugins/vocord/

3. Set up backend:

   macOS ARM (mlx-whisper):

   uv venv ~/.local/share/vocord/venv --allow-existing
   uv pip install --python ~/.local/share/vocord/venv/bin/python mlx-whisper
   echo "mlx-whisper" > ~/.local/share/vocord/backend

   Linux / Windows / macOS Intel (transcribe-rs):

   # ffmpeg (for ogg->wav conversion)
   brew install ffmpeg          # macOS
   sudo apt install ffmpeg      # Debian/Ubuntu
   winget install ffmpeg        # Windows
   
   # Build and install transcribe-cli
   cd /tmp/vocord/transcribe-cli
   cargo build --release
   cp target/release/transcribe-cli ~/.local/share/vocord/
   
   # Download Whisper large-v3-turbo GGML model (~800 MB)
   curl -L --fail -o ~/.local/share/vocord/ggml-large-v3-turbo.bin \
     https://blob.handy.computer/ggml-large-v3-turbo.bin
   
   echo "transcribe-rs" > ~/.local/share/vocord/backend

4. Build:

   cd /path/to/Vencord && pnpm build

5. Connect your client:

   • Vesktop: Settings > Developer Options > Vencord Location → select Vencord/dist/
   • Discord Desktop: cd /path/to/Vencord && pnpm inject

6. Restart Discord / Vesktop, enable Vocord in Settings > Vencord > Plugins.

Architecture

Discord Voice Message
  |  click
  v
index.tsx (browser) ---IPC---> native.ts (Node.js)
                                  |
                         reads ~/.local/share/vocord/backend
                          /             \
                    macOS ARM64      everything else
                    (mlx-whisper)         |
                         |           ffmpeg (ogg->wav)
                         |                |
                   python venv       transcribe-cli
                  (whisper-large-     (Whisper GGML)
                    v3-turbo)
                          \             /
                           JSON result
                         {"text": "..."}

Troubleshooting

Error	Fix
Plugin not showing (Vesktop)	The installer auto-configures Vesktop. If it didn't work: Vesktop Settings > Developer Options > Vencord Location → select your Vencord dist/ folder
Plugin not showing (Discord)	Rebuild (cd Vencord && pnpm build) and inject (pnpm inject), then restart Discord
mlx-whisper not found	Re-run the installer, or: uv pip install --python ~/.local/share/vocord/venv/bin/python mlx-whisper
ffmpeg not found	brew install ffmpeg / sudo apt install ffmpeg
transcribe-cli not found	Re-run the installer to rebuild the binary
Whisper model not found	Re-run the installer to download the model

License

GPL-3.0-or-later

Credits

• Vencord
• mlx-whisper / mlx-community/whisper-large-v3-turbo
• transcribe-rs
• Handy

Author

jolehuit