Free AI voiceovers on your laptop, no ElevenLabs bill
Natural-sounding narration for videos, tutorials and audiobook drafts, made by the open Kokoro model on an ordinary laptop. No subscription, no account, no graphics card.
- 28 English voices, British and American, and the trick for fixing words it says wrong
- Level 0 helps: you'll use the terminal and uv (step 1 installs it if you don't have it)
- About 650MB of free disk space
- 8GB of RAM is plenty. No graphics card needed
Make a folder and install Kokoro
uv keeps everything inside one folder, so it can't clash with anything else on your PC and is easy to delete later.
Hear what you're about to make, all generated on a laptop with no graphics card: a British voice and an American one.
Kokoro is a small, open text-to-speech model (82 million parameters, tiny by AI standards). The model weights are Apache-2.0, which means you can use the audio in commercial videos. You run it with kokoro-onnx (MIT licence), a lightweight package that works on any normal CPU.
If you don't have uv yet, install it first, then open a new terminal:
Windowswinget install --id astral-sh.uv -eMac or Linux (or `brew install uv` on a Mac with Homebrew)curl -LsSf https://astral.sh/uv/install.sh | shNow make a project folder and install Kokoro into it:
Terminalmkdir kokoro-voices cd kokoro-voices uv init -p 3.12 uv add kokoro-onnx soundfileOutputInitialized project `kokoro-voices` Resolved 17 packages in 629ms ... Installed 16 packages in 13.08s + kokoro-onnx==0.6.1 + numpy==2.5.3 + onnxruntime==1.30.0 + soundfile==0.14.0 ...Tip-p 3.12tells uv to use Python 3.12 for this folder. If you don't have it, uv downloads it for you. It doesn't touch any other Python on your PC.
kokoro-onnx and soundfile with a + in front.