Get started

Single interactive network installer, works for a fresh install or an upgrade. It auto-detects GPU acceleration (CUDA, Vulkan, Metal or CPU) and installs the most suitable build.

curl -fsSL https://raw.githubusercontent.com/DavidValin/vtmate/main/installer.sh | sh

wget -qO- https://raw.githubusercontent.com/DavidValin/vtmate/main/installer.sh | sh

Compatible with:

macOS

Windows

Linux

Or grab a release by hand from the releases page, install it via cargo, or build it from source.

Install an LLM engine

If you need ai responses vtmate needs somewhere to send the transcribed text for a reply. Pick one:

curl -fsSL https://ollama.com/install.sh | sh
ollama pull llama3.2:3b

Optional: OpenTTS support (an extra TTS voice engine)

docker run --rm --platform=linux/amd64 -p 5500:5500 synesthesiam/opentts:all

Acceleration

For CUDA, install the CUDA Toolkit (12.x or 13.x) and cuDNN 9 (on Windows, put cuDNN's bin\<cuda major>.x directory on PATH, or copy its DLLs next to the other vtmate libraries). installer.sh checks both are reachable and falls back to the Vulkan/CPU build otherwise. For Vulkan, install the Vulkan SDK.

Building from source

For a full configurable build (OS, architecture and GPU acceleration), use the platform scripts below. For a quick CPU-only build instead:

CPU only

cargo build --release

Full configurable builds (OS, architecture and GPU acceleration)

build_linux.sh
build_macos.sh
build_windows.ps1

Troubleshooting

If vtmate won't start, remove ~/.vtmate/settings and/or ~/.vtmate/agents so it recreates the default configuration on the next run. Full option reference:

vtmate --help