Installation

Installation

On Windows 11 you'll be able to install the Windows app instead, with no developer tools: it comes with version 2.1.0, which isn't published yet. This page covers running TTSNative from its source code (Windows, macOS and Linux), which is how you run it today and how you develop it.

You get	Windows app	From source
Python, server, voice model	Included in the installer	Downloaded by the first npm run dev
Updates	Offered in the app, signature-checked	git pull
Your data	%LOCALAPPDATA%\com.ttsnative.desktop\data	data/ in the project folder
Usual port	8770 (up to 8789 if busy)	8765 for the app, 5173 for npm run dev's page
Phones (LAN mode)	Yes	With npm run start, not npm run dev

Requirements

You need Node.js 22.22 or newer and Git. You don't need to install Python yourself: the first run installs everything else inside the project folder.

	Basic reader (Kokoro)	Voice cloning (Chatterbox, optional)
Disk	~0.7 GB (runtime + model)	+8 GB with NVIDIA CUDA (measured); about +4.5 GB CPU-only (estimated)
Memory	~1 GB while speaking	4 GB+ RAM; 4 GB+ VRAM with a GPU
GPU	Not needed	Recommended (NVIDIA or Apple Silicon); CPU works but is slower than real time
Network	First run only	One-time download from Settings

CI runs the server's tests on Windows, macOS and Linux, and the web and browser tests on Linux. The full app hasn't been tested by hand on macOS or Linux; see the FAQ.

Install

git clone https://github.com/carsonmcrae/ttsnative.git
cd ttsnative
npm install
npm run dev

When the app is ready, your browser opens at http://127.0.0.1:5173. Press Ctrl+C in the terminal to stop it. To start it again, run npm run dev in the same folder; later starts take a few seconds.

If port 5173 (web) or 8765 (API) is busy, the app picks the next free port and prints the address.

What the first run does

1. Downloads a pinned copy of uv, the Python manager, and checks its size and SHA-256 checksum. If uv 0.9.0 or newer is already on your PATH, that copy is used instead (set LTTS_UV_PINNED_ONLY=1 to always use the pinned download).
2. uv installs Python 3.12 into .tools/ and the locked server dependencies into apps/server/.venv. No system Python is required or changed.
3. Downloads the Kokoro voice model (about 340 MB) and verifies its size and SHA-256 before installing it.

If the model can't be downloaded (for example, you're offline), the app still starts and shows Install the voice model to start listening. Install it from that banner, from Settings → Engines & models, or by running npm run setup once you're online. Setting LTTS_SKIP_MODEL_DOWNLOAD=1 skips the download on purpose.

Where things are stored

Everything lives inside the project folder and is ignored by Git.

Location	Contents
.tools/	uv and Python 3.12
apps/server/.venv	The server's Python environment
data/ltts.sqlite3	Your library, reading progress, preferences and voice details
data/models/	Verified model files
data/engines/	Optional engine environments (Chatterbox)
data/cache/audio/	Generated speech, reused when you listen again
data/voices/	Voice Studio reference recordings
data/exports/	Exported audio

Settings → Storage shows the data folder and how much space each part uses. To keep your data somewhere else, set LTTS_DATA_DIR (see Settings and Models). To back it up, stop the app and copy data/. The Windows app keeps its data elsewhere; see Where your data is.

Production-style run

npm run start builds the web app and serves the UI and the API from one process, at http://127.0.0.1:8765 (or the next free port). npm run dev is the everyday command and adds hot reload for development.

Updating

TTSNative is updated with Git:

git pull
npm install
npm run dev

Every start re-syncs the Python environment from the committed lockfile, so new server dependencies are installed automatically. Your data/ folder isn't tracked by Git, so your library, voices and models are kept.

The changelog and the releases page list what changed in each version, and Settings → About shows the version you're running.

Uninstalling

Nothing is installed system-wide. To remove TTSNative, stop the app (Ctrl+C) and delete the project folder. If you pointed LTTS_DATA_DIR, LTTS_MODELS_DIR or LTTS_ENGINES_DIR at a folder outside the project, delete that folder too.

To start over without cloning again, stop the app and delete data/ (your library, recordings and models) and .tools/ (uv and Python), then run npm run dev.

To free space without uninstalling, remove an engine or clear the audio cache in Settings.

Checking your setup with npm run doctor

npm run doctor

The doctor checks Node, uv, the Python environment, installed models, GPUs, ports and disk space. It only reports: it never installs or changes anything. Run it first when something doesn't work, and include its output (or Settings → About → Copy diagnostics) when you report a bug. See also Troubleshooting.