H3 FL2VA vs H3-Image character knowledge
retained text
Testing character knowledge in H3-image vs the original H3 FL2VA model. Same sampler settings and seed for each image. LCM sgm-uniform, seed 34, 8MP. First image is always H3 FL2VA and second image the new H3-image model. 1. Sheldon from The Big Bang Theory. 2. Penny from The Big Bang Theory. 3. Harry Potter. 4. Jack Sparrow.
document fetched 2026-11-12 00:32 · canonical https://reddit.com/r/StableDiffusion/comments/1xtiv2z/h3_fl2va_vs_h3image_character_knowledge