Oleg Oshmyan

> If that's implemented as an `ASS_FEATURE`, do we really want font fallback to be disabled by `INCOMPATIBLE_EXTENSIONS`? Doing it as an `ASS_Feature` does seem sensible, and no, I imagine...

First: when `Language` is set, I agree we should take it into account for font fallback (I’ve thought about this myself). When it isn’t: VSFilter’s effective font fallback seems to...

Tangential note: it would also be good to provide a language _override,_ because a single track can contain mixed-language text. But the usual caveats about adding new overrides apply.

Hmm, this is not strictly related to language issues, but do you have any idea where your `Source Han Sans` comes from at all? Judging by [their libass Makefile](https://code.videolan.org/videolan/vlc/-/blob/32b04e93bc43e1b4a75f6fa5d52fb70613dc9507/contrib/src/ass/rules.mak#L10-11), Android...

TODO: retest font fallback in VSFilter with the style’s `Encoding` (or `\fe`) set to things other than 1. I suspect any other valid `Encoding` may define which locale the fallback...

@rcombs Do you remember where this ended up at? This PR’s discussion thread dove deeper into font matching, and I gather some of the issues mentioned have been resolved by...

The current architecture of ASS and of libass does not permit multicoloured glyphs. Theoretically, we could support colour emoji in the very simplest case of opaque plain text; but when...

Yes! I’ve said the same thing before, but it requires another fairly major change to the code, so there is no telling when we may get it done. Rasterizing whole...

> also, as aside note, assrender don't load the font that inside the .ass file unlike vsfilter which did load it, it is problem in assrender or libass itself? Most...

As mentioned on IRC, the 32-bit MinGW-w64 build is failing because GCC seems to be mistakenly assuming the stack is pre-aligned on a 8-byte (or more) boundary, which it isn’t,...