v1.4.0 - Fixes and performance improvements ⏩
·
49 commits
to main
since this release
Download
This release is available via the Creator Companion as a VPM package!
⬇️ Creator Companion/VPM Install
For standalone Unity, you can import LTCGI as a git package with the URL: https://github.com/PiMaker/ltcgi.git
Adapters
LTCGI adapters are now built-in. Migration should be automatic. You do not need to download anything specific, having USharpVideo or AvPro installed in your project should automatically enable the corresponding adapters.
Double-check your adapters after updating - there is a chance you may need to reconfigure them in your scene, depending on which version you upgraded from.
APIv2
This contains the first version of the v2 API. Read more here: https://ltcgi.dev/Advanced/Shader_Authors
Fixes
- Fix warning spam
X4000 use of potentially uninitialized variableby using out variables - thanks @ScruffyRules! - Always create shader redirect file from Assets to Packages, not just when controller is placed in scene
- Fix an issue where single-UV sampling would use the wrong position - it now always uses the exact position on mip0 of the video texture
- Add 2 performance-related controller options (available in the "advanced settings" part of the LTCGI_Controller, note that these will be reset every update):
LTCGI_BLENDED_DIFFUSE_SAMPLING: slightly improves diffuse quality for an extra texture tap (perf) - current versions always do that, for performance it is now behind a default-off flag as the effect was very minimalLTCGI_FAST_SAMPLING: saves a few texture slots and improves performance by using no prefiltering and just using the raw mip-chain of input textures for sampling blurred versions - noticeable reduces specular quality, but has a slight impact on perf (not huge, but measurable) - I recommend considering hardcoding this define for avatar shader authors, or trying it in your world to see if it makes a big visual impact and leaving it on if not
- Generally improve performance of UV sampling - behaviour should be unchanged, but I noticed about a 10% raw throughput improvement for textured reflections (stress-tested with 128 screens on LTCGI_Surface, 7900 XTX; I do not recommend using more screens than absolutely necessary even with this improvement)
Contributors
ScruffyRules