Write Custom Filters

A Vapourkit filter is a snippet of VapourSynth Python that transforms clip into a new clip. The filter runs inside Vapourkit’s processing pipeline alongside AI upscaling.

Minimum example

# Inside a filter, `clip` is the input and you reassign it to apply changes.
clip = core.std.Crop(clip, left=4, right=4, top=0, bottom=0)

Available APIs

Anything core.* from VapourSynth and any plugin loaded by Vapourkit is fair game. Loaded plugins include vszip, vs-mlrt, vs-jetpack, and many others — see the existing Filter Reference for examples of plugin usage. The loaded set depends on the platform: Windows includes the complete bundled catalog, while Linux includes only plugins verified by its PyPI-based setup. A custom filter that imports an unavailable module or native plugin will fail when previewed or processed.

Filter metadata

When you save a filter as a template, Vapourkit serializes it to .vkfilter (TOML) with these fields:

• name — display name in the UI
• category — used for grouping in the filter picker
• code — the Python snippet
• description — short summary shown on hover
• [metadata] block — timestamps and bookkeeping

See File Formats for the schema.

Sharing

Filter templates live in data/config/filter-templates/ by default. On Linux this is under the AppImage’s external application-data directory, not inside the read-only AppImage. You can email or paste templates — they are plain text TOML — but the recipient still needs every referenced plugin and Python module installed on their platform.