Guides
Start with the quickstart, then open only the guide you need. The four port
guides include their own download and build steps. Python snippets assume
pip install "loudkit[torch,audio,hub]" and weights reachable by repo id or by
a checkpoint file at loudr-1.safetensors. See
the quickstart for how to get both.
| # | Guide | What you will have when it is done |
|---|---|---|
| 1 | Getting started | Your first WAV, from a two-line Python API |
| 2 | Streaming and long-form | First audio before the passage finishes; long texts split across windows |
| 3 | Cloning a voice | A voice profile that is yours, from ten seconds of audio |
| 4 | Server, streaming API and MCP | Any local app or agent can speak |
| 5 | Benchmarking and profiling | A reproducible number for your hardware, stage by stage |
| 6 | Embedding loudkit | loudkit inside your own program, on the right device |
| 7 | JavaScript / TypeScript | the same engine from Node, over onnxruntime-node |
| 8 | Go | the same engine from Go, over yalue/onnxruntime_go |
| 9 | Rust | the same engine from Rust, over the ort crate |
| 10 | Swift | the same engine from Swift, over CoreML, with the estimator on the Neural Engine |
The documentation index links the API and implementation reference. Guides 1 through 6 use Python; guides 7 through 10 cover the other four implementations.