Optional multiline reference_list: one base64 image per line, "[strength ]<base64>".
Image count = non-empty lines, so a single static graph adapts to any number of
references with no empty slots, no placeholder images, no rewiring. Decodes
base64->IMAGE internally. Works alongside image_N sockets; normalize/strength=0
apply across both. README updated, version 1.2.0.
normalize scales each active image's strength to sum to 1, preserving
relative ratios (2/1/1 -> 0.5/0.25/0.25; all equal -> 1/N). strength=0 still
skips in both modes. Default stays manual. README + version bumped to 1.1.0.
Rename FUNCTION apply→inject_references, drop the if!=1.0 scale special-case,
tighten strength to [0,5], rewrite all comments. README: honest strength
"approximation + limits" section, VRAM/resolution note, fix range to 0-5.
Independent of any prior reference-latent extension.
Inject up to 6 reference images into conditioning for FLUX.2 (and other
edit models consuming reference_latents) in a single node. Each image has
its own strength (0 = skip), built-in VAE encode, all inputs optional so
one static workflow covers 0–6 references without bypass or rewiring.
Independent implementation based on ComfyUI's public native ReferenceLatent
mechanism.
lynclee