3 ms·
SDXL(and possibly 2.1) switched to different CLIP implementation that is geared for sentence-level understanding, SD1.5 uses old CLIP that works with tag-cloud
by countWSS 3y ago
SDXL(and possibly 2.1) switched to different CLIP implementation that
is geared for sentence-level understanding,
SD1.5 uses old CLIP that works with tag-cloud type prompts.
- ttul 3y agoSDXL actually takes conditioning from either the old or the new CLIP, or both. The malleability of SDXL is not just down to the choice of the new CLIP; the UNet itself is more opinionated.