Nodes/ComfyUI_agilly1989_motorway/VideoLinearCFGGuidance_motorway_edition
ComfyUI Node

VideoLinearCFGGuidance_motorway_edition

Ease CFG in over the first frames instead of hitting it all at once

By agilly1989·Created 2 years ago·Updated about a year ago· 7
VideoLinearCFGGuidance_motorway_edition
  • MOTORWAY 🚌💨
  • MOTORWAY 🚌💨
INPUT_model_keymodel
min_cfg1.0
OUTPUT_MODEL_keyMODEL

Full CFG from the very first frame is often the wrong move for video. Video models like SVD want guidance to ramp in - start loose, tighten up as the clip develops - which is exactly what VideoLinearCFGGuidance does: it patches the model so CFG scales linearly from a minimum on the first frame up to your sampler's full CFG on the last. This _motorway_edition clone applies that same patch through the agilly1989 Motorway pipe, pulling the model from the shared dict by key.

Motorway primer: start with MotorwayStart, then each _motorway_edition node trades data wires for key strings, reading and writing a shared dict that flows down one MOTORWAY connection. These nodes are auto-generated clones of core ComfyUI nodes, so the patching logic is identical to stock.

What it does. It clones your MODEL and installs a custom CFG function. During sampling, instead of uncond + cond_scale * (cond - uncond) applied uniformly to every frame, the scale interpolates from min_cfg on the first frame up to the full cond_scale on the last. The only knob is min_cfg (default 1.0, range 0–100) - how weak guidance is at the start. Lower values mean the early frames are more free-form, which is the classic recipe for stable, natural motion; the sampler's own CFG value is still the ceiling. This is the kind of node you'd slot between the model load and the sampler when a video workflow produces jittery or oversaturated early frames.

The inputs that matter. MOTORWAY is required. INPUT_model_key (default model) is where the base model lives in the dict - and the case trap applies, since loaders store under MODEL uppercase by default. min_cfg is the only creative decision. OUTPUT_MODEL_key (default MODEL) is where the patched model is stored for the sampler. Output: the MOTORWAY port only.

Common issues. First, this only matters for video models with a temporal dimension, and it's a guidance tweak, not a quality guarantee - pair it with sensible sampler settings or it'll do nothing visible. Second, the Motorway key/case traps: 'MotorwayClass' object has no attribute 'hash_' means a key wasn't stored, so line up MODEL vs model before debugging anything else. And the standing caveat: the pack's current main branch has the clone registration commented out in __init__.py, so if the node's missing from your menu, that's the version you're on.

Install via ComfyUI Manager (search "ComfyUI_agilly1989_motorway") or clone into custom_nodes, restart, no models to download. Solo-dev beta pack - the Motorway is a tidy-up layer, and for video sampling chains with several stages, having the guidance patch live in the same dict as the model is a legitimately convenient fit.

Categoryagilly1989 Nodes/Motorway-ed/sampling/video_models

Inputs (4)

NameTypeDefaultDescription
MOTORWAY 🚌💨MOTORWAY 🚌💨
INPUT_model_keySTRINGmodel
min_cfgFLOAT1.00–100
OUTPUT_MODEL_keySTRINGMODEL

Outputs (1)

NameTypeDescription
MOTORWAY 🚌💨MOTORWAY 🚌💨