Installation ​

Recommended Installer ​

curl -fsSL https://ostt.ai/install | bash

Use this unless you specifically prefer a platform package manager. It detects your platform, installs missing runtime dependencies (ffmpeg, clipboard tools) where supported, selects the right CPU/GPU build, downloads the latest release, verifies its checksum, and installs the ostt CLI. On Omarchy 4, it also configures ALT+SPACE for clipboard dictation, ALT+SHIFT+SPACE for spoken prompts to the default agent, and the floating OSTT popup. In most cases you do not need to install dependencies or desktop integration manually.

The installer runs non-interactively. When no transcription model is active at the end of the install, it offers to download the local Whisper base model (142 MB) so you can start transcribing immediately -- accepting the offer leaves a fully working setup with nothing else to configure.

It uses native .deb or .rpm packages on supported Linux distributions when possible. Otherwise, including macOS, it installs the binary to ~/.local/bin by default. If that directory is not already on your PATH, the installer adds it to your shell profile (.bashrc, .zshrc, or config.fish) so ostt works in your next shell with no manual steps. Pass --no-modify-path to opt out and get printed instructions instead.

Shell completions are installed to user-owned directories -- no sudo required. For zsh, the installer also adds the completions directory to fpath in your .zshrc, since zsh's default fpath contains no directory under $HOME.

Download URLs and checksums come from a release manifest at ostt.ai/latest.json, so the installer never guesses filenames. If the manifest is unreachable the install fails with an explicit error rather than falling back to a guess. Installing a pinned release with --version bypasses the manifest, since it only ever describes the latest release.

On macOS, the installer uses the Metal-enabled build. On Linux x86_64, it installs the CUDA build when a usable NVIDIA CUDA runtime is detected, the Vulkan build when AMD/Intel Vulkan support is detected, and the CPU build otherwise. To force the CPU build:

curl -fsSL https://ostt.ai/install | bash -s -- --no-gpu

Installer options ​

Pass options after -s --. For example, an install that keeps the CPU build and leaves your shell profile alone:

curl -fsSL https://ostt.ai/install | bash -s -- --no-gpu --no-modify-path

The install directory can also be set with the OSTT_INSTALL_DIR environment variable. When --install-dir is left at the default and a native .deb/.rpm package is available for your system, the installer uses that package (installing to /usr/bin) instead of the release archive.

First Run ​

OSTT requires an active transcription model before recording. When none is active, the installer ends by offering to download the local Whisper base model (142 MB) and selecting it for you. If the base model is already on disk from an earlier install, it is selected without asking. Accept the offer and OSTT is ready to transcribe -- no further setup needed.

If you decline (or ran the installer with --no-model-download or without a terminal), pick a model yourself:

ostt model

You can select and download a local model directly. To use a cloud provider, authenticate it first and then select one of its models:

ostt auth login
ostt model

On Omarchy, also select the coding agent used by the voice-to-agent hotkey if you have not already done so:

omarchy agent --pick

The Omarchy installer changes are append-only and idempotent. Existing bindings, OSTT window rules, and [process.actions.agent] definitions are preserved instead of overwritten. See Omarchy / Hyprland Setup for the generated configuration and manual setup.

Native Package Options ​

macOS -- Homebrew ​

brew tap kristoferlund/ostt
brew install ostt

Arch / Omarchy -- AUR ​

Prebuilt binary packages (recommended -- no compilation, no Rust toolchain). Pick the one that matches your hardware:

yay -S ostt-bin          # CPU build (x86_64, aarch64)
yay -S ostt-cuda-bin     # NVIDIA CUDA build (x86_64)
yay -S ostt-vulkan-bin   # AMD / Intel Vulkan build (x86_64)

These packages download the official release binary and install it directly. They conflict with each other and with the source package, so only one can be installed at a time.

To build from source instead:

yay -S ostt

paru works in place of yay for any of the above.

Debian / Ubuntu / Mint -- .deb ​

# x86_64 CPU build
curl -sLO https://github.com/kristoferlund/ostt/releases/latest/download/ostt_latest_amd64.deb
sudo apt install ./ostt_latest_amd64.deb

# x86_64 NVIDIA CUDA build
curl -sLO https://github.com/kristoferlund/ostt/releases/latest/download/ostt-cuda_latest_amd64.deb
sudo apt install ./ostt-cuda_latest_amd64.deb

# x86_64 AMD/Intel Vulkan build
curl -sLO https://github.com/kristoferlund/ostt/releases/latest/download/ostt-vulkan_latest_amd64.deb
sudo apt install ./ostt-vulkan_latest_amd64.deb

# ARM64 (Raspberry Pi, etc.)
curl -sLO https://github.com/kristoferlund/ostt/releases/latest/download/ostt_latest_arm64.deb
sudo apt install ./ostt_latest_arm64.deb

Fedora / RHEL -- .rpm ​

# x86_64 CPU build
sudo dnf install https://github.com/kristoferlund/ostt/releases/latest/download/ostt-latest.x86_64.rpm

# x86_64 NVIDIA CUDA build
sudo dnf install https://github.com/kristoferlund/ostt/releases/latest/download/ostt-cuda-latest.x86_64.rpm

# x86_64 AMD/Intel Vulkan build
sudo dnf install https://github.com/kristoferlund/ostt/releases/latest/download/ostt-vulkan-latest.x86_64.rpm

# ARM64
sudo dnf install https://github.com/kristoferlund/ostt/releases/latest/download/ostt-latest.aarch64.rpm

openSUSE -- .rpm ​

# x86_64 CPU build
sudo zypper install https://github.com/kristoferlund/ostt/releases/latest/download/ostt-latest.x86_64.rpm

# x86_64 NVIDIA CUDA build
sudo zypper install https://github.com/kristoferlund/ostt/releases/latest/download/ostt-cuda-latest.x86_64.rpm

# x86_64 AMD/Intel Vulkan build
sudo zypper install https://github.com/kristoferlund/ostt/releases/latest/download/ostt-vulkan-latest.x86_64.rpm

# ARM64
sudo zypper install https://github.com/kristoferlund/ostt/releases/latest/download/ostt-latest.aarch64.rpm

Run uname -m to check your architecture (x86_64 or aarch64). CUDA and Vulkan packages are currently only published for Linux x86_64. Use the CPU package on ARM64.

Choose the package variant by hardware:

Direct Binary Download ​

Download from GitHub Releases:

# CPU archives: x86_64-linux, aarch64-linux, x86_64-macos, aarch64-macos
# GPU archives: x86_64-linux-cuda, x86_64-linux-vulkan
tar -xzf ostt-<platform>.tar.gz
sudo cp <extracted-path>/ostt /usr/local/bin/

Compile from Source ​

git clone https://github.com/kristoferlund/ostt.git
cd ostt
cargo build --profile dist --locked
# Binary at target/dist/ostt

Runtime Dependencies ​

macOS ​

brew install ffmpeg       # ffmpeg
# pbcopy is built-in for clipboard
brew install mpv          # Optional: better audio playback

Debian / Ubuntu ​

sudo apt install ffmpeg wl-clipboard   # Wayland
# OR
sudo apt install ffmpeg xclip          # X11
sudo apt install mpv                   # Optional: better audio playback

Arch / Omarchy ​

sudo pacman -S ffmpeg wl-clipboard     # Wayland
# OR
sudo pacman -S ffmpeg xclip            # X11
sudo pacman -S mpv                     # Optional: better audio playback

Fedora ​

sudo dnf install ffmpeg wl-clipboard   # Wayland
# OR
sudo dnf install ffmpeg xclip          # X11
sudo dnf install mpv                   # Optional: better audio playback

Native package managers (Homebrew, AUR, .deb, .rpm) declare these dependencies and install most of them automatically.