The libvips CLI usually executes a pipeline of three operations: a loader (selected from the magic number in the input data), the operation named first in the cli arg, and a saver selected from the output filename extension.

Any --flag xxx options are sent to the named operation, but that leaves no way to pass options to the loader and saver, so libvips uses the strange [x=y] syntax on filenames for this.

vips gravity x.jpg[shrink=2] y.tif[compression=lzw] centre 1000 1000 --background "255 0 0"

Executes something like:

jpegload x.jpg --shrink 2 | 
    gravity centre 1000 1000 --background "255 0 0" | 
        tiffsave y.tif --compression lzw

Having the load and save options in the filenames is often very convenient. You can write a little script to do something and not have to worry about file formats at all, just use $1 or whatever. The person using your script can process GIF or AVIF or whatever they like just by adding filename options.

this vips heifsave src test --compression=av1 --encoder=aom works only for this use case, if i want to perform any kind of action, like flip(vipsflip src flip_y VIPS_DIRECTION_VERTICAL --compression=av1 --encoder=aom) it errors with unknown arguments.

right now i have to use file extension and then rename the file afterwards.

If you write to a suffix with no filename, it'll write to stdout. So you could do:

vips flip xxxx .avif[compression=av1,encoder=aom] horizontal > yyyy

i might use that in the future, but for now using the renaming seems the way to go.