Troubleshooting & FAQ
This page is a work in progress. If you hit something that isn’t covered here, ask in the Discord or open an issue on GitHub.
Setup fails
If the first-launch setup fails partway through, restart Vapourkit and setup will resume from where it left off. The plugins-and-filters stage can also be re-run later from the Plugins menu in the header without losing other components.
TensorRT errors
Make sure your NVIDIA driver is at least version 580.x. Older drivers won’t load the bundled TensorRT runtime.
DirectML is slow
DirectML is more compatible but slower than TensorRT. If you have a supported NVIDIA GPU, prefer TensorRT.
Out of memory
If processing fails with an out-of-memory error:
- Add a Resize filter before the model to lower the working resolution.
- Close other GPU-heavy applications.
- Try DirectML instead of TensorRT if you’re on NVIDIA — it has different memory characteristics.
Missing plugins
If filter previews fail with errors about missing VapourSynth functions, the plugin install may have been skipped or interrupted. Reinstall plugins from the Plugins menu in the header.