Cutoff Regions To Conditioning (ADV)

This is the "advanced" version of Cutoff Regions To Conditioning, and "advanced" here means exactly two extra dropdowns: token_normalization and weight_interpretation. If you're just trying to get "blue hair, yellow eyes, red shoes" to render with the right colors on the right nouns, you don't need this node - the plain one does the same job with saner defaults. Reach for the ADV variant when your prompt uses weighting syntax like (detail:1.3) and the base node's treatment of those weights is giving you mushy results.

Those two knobs are borrowed wholesale from BlenderNeko's other pack, ComfyUI_ADV_CLIP_emb - the advanced CLIP text encoder that popularized this exact set of weight-handling options. In fact, both finalizer nodes call the same underlying function; ADV just passes two extra parameters through.

What the extra inputs do

• token_normalization - none, mean, length, length+mean. Controls how raw prompt weights are normalized across the chunk before encoding. mean shifts weights around their average, length scales by the number of tokens a word spans. Think of it as "how literal should a weight be."
• weight_interpretation - comfy (default), A1111, compel, comfy++. This is the big one: it's which prompt-weighting dialect your (word:x) and [word:y] syntax gets interpreted as. If you're porting a prompt from A1111 or the Compel library, pick the matching mode so the weights mean what the original author intended.

Everything else - clip_regions, mask_token, strict_mask, start_from_masked - is identical to the non-ADV node, and the output is a normal CONDITIONING that drops into your KSampler's positive input.

Honest advice

The plain Cutoff Regions To Conditioning node defaults to comfy interpretation, which is what ComfyUI-native prompts already use. The ADV node earns its place in two situations: you're feeding it prompts with heavy weighting that comes out weak or oversaturated, or you're migrating an A1111 workflow and want A1111-style weighting respected inside your regions. Otherwise the extra dropdowns are just noise - set them wrong and you'll spend an hour chasing a subtle quality difference that's the weighting dialect, not your regions.

Install

ComfyUI Manager (search "Cutoff"), or:

cd ComfyUI/custom_nodes
git clone https://github.com/BlenderNeko/ComfyUI_Cutoff

Restart ComfyUI. No dependencies, no model downloads - the whole pack is two Python files, and the ADV encode helpers ship right inside it, so you don't need the ADV_CLIP_emb pack installed separately for this node to work. Find it under conditioning > cutoff, next to its simpler sibling.