SnapRef

Preview (step 15,000 of the v2a run). Try this one first among the previews. The new Character mode takes the character from the picture and leaves its art style behind, so the base model and your prompt set the style. In Character &amp; Style mode it is close to v1a but not better; v1a stays the recommended version for everyday use.LoRA rank 128 / 64, 382 MB (FP8 195 MB). Needs the SnapRef node from the attached ZIP (a LoRA loader will not work). Open <code>workflows/snapref_v2a_pre2_simple.json</code>, pick the picture inside the SnapRef node and write only the scene.

Files with SHA256: 1c9794ac626551654aa0bf08028e70a451eda4229878a7ea00490b4cc72edaa8