Custom ONNX Models

Vapourkit supports any model that vs-mlrt supports. To use your own:

1. Place the .onnx file in data/models/ inside your Vapourkit install directory.
2. Restart Vapourkit (or reload models from the menu).
3. Your model appears in the model picker.

Supported architectures

Anything vs-mlrt accepts. See vs-mlrt model documentation.

TensorRT engines

For NVIDIA users, Vapourkit can compile your ONNX model into a TensorRT engine (.engine file) for faster inference. This step is GPU- and driver-specific — you can’t share .engine files across machines.

Finding models

OpenModelDB is the most active community catalog.

Licensing

Check the license of any model you import — most community models are CC BY-NC-SA. See Licensing.