Nodes/IAMCCS-nodes/IAMCCS Cine Temporal Cut Barrier (Experimental)
ComfyUI Node

IAMCCS Cine Temporal Cut Barrier (Experimental)

Forcing hard cuts inside a single generation (experimental, obviously)

By IAMCCS·Created 11 months ago·Updated 7 days ago· 113
IAMCCS Cine Temporal Cut Barrier (Experimental)
  • model
  • cine_linx
  • model
  • report
enabledfalse
boundary_modemarked_cuts
barrier_strength1.00
first_block0
last_block-1

Most video pipelines fake cuts the boring way: generate clip A, generate clip B, splice in the editor. Temporal Cut Barrier is the pack's attempt at something bolder - making a single generation contain an actual hard cut, by patching the model so the timeline treats a cut boundary as a wall that earlier frames can't leak across. It's marked experimental in the category name, and it ships disabled by default. Read that as a design statement.

How it works. It takes your model and the cine_linx plan, and if enabled is on, it patches the model with attention/block barriers at the cut boundaries the plan marks. boundary_mode picks where the barriers go: marked_cuts uses only the boundaries you explicitly marked as cuts in the shotboard, while every_visual_slot places a barrier at every visual segment boundary. barrier_strength (0–1, default 1) controls how hard the wall is, and first_block/last_block (defaults 0 and -1) restrict the barrier to a block range - last_block = -1 meaning "through the end." The output is the patched model and a report string describing what was applied.

The practical promise: a hard cut mid-generation instead of a mush where the model tries to morph one shot into the next. The experimental caveat: this is patching the model's internals to do something the model wasn't trained to express, so results vary - that's exactly why enabled defaults to false and the category says experimental.

The inputs. model, cine_linx, enabled (bool, off), boundary_mode, barrier_strength, first_block, last_block. When enabled is off, the node is a pure passthrough - the report literally says "Temporal Cut Barrier: OFF (model unchanged)."

Where it fits. Between your backend and your sampler, as a model patch. It's a Stage-1 experiment per the source ("Optional single-generation hard-cut experiment for Stage 1"), so think of it as an option to try when continuity across a marked cut keeps failing and you want the cut to be a hard break rather than a soft drift.

Install. Part of IAMCCS-nodes: ComfyUI Manager (search "IAMCCS"), or git clone https://github.com/IAMCCS/IAMCCS-nodes into custom_nodes, restart. No models, no extra deps.

The honest take. Approach this one with expectations set low and A/B habit on: run the same shot with enabled on and off, compare, keep whichever wins. The barrier can clean up a cut, and it can also introduce a jarring discontinuity in motion - that's the trade you're signing up for. If you're not hitting a specific cut problem, there's no reason to enable it. Also check the report when results look weird: it tells you the boundary mode and block range that actually got applied, so a mismatch between what you set and what you intended is visible in one string.

CategoryIAMCCS/Cine/03 Experimental

Inputs (7)

NameTypeDefaultDescription
modelMODEL
cine_linxIAMCCS_SUPERNODE_LINX
enabledBOOLEANfalse
boundary_modeCOMBOmarked_cuts2 options: marked_cuts, every_visual_slot
barrier_strengthFLOAT1.000–1
first_blockINT00–100
last_blockINT-1-1–100

Outputs (2)

NameTypeDescription
modelMODEL
reportSTRING