Visualize Context Options (S.Cus.) 🎭🅐🅓

The third member of the Visualize Context Options family, and the one for people who've already left the beaten path of sampler_name + scheduler. If your AnimateDiff pipeline builds its own SIGMAS schedule - a custom scheduler node, a hand-tuned curve, anything that isn't a stock sampler/scheduler combo - the other two visualize nodes can't reconstruct it for you. This one skips the reconstruction entirely and takes your actual sigmas as input, so the preview is built from exactly the schedule you're really going to sample with.

How it works

Same underlying idea as its two siblings: no pixels, no motion module execution, just a simulation of which frames land in which context window at each sampling step, rendered as a chart. The difference is what it's stepping through - instead of deriving a timestep schedule from a sampler/scheduler pair, it walks your sigmas tensor directly. That makes it the more accurate choice whenever your real pipeline's schedule didn't come from a plain KSampler node in the first place.

The inputs and outputs that matter

• model (MODEL, required) - the model you're sampling with.
• sigmas (SIGMAS, required) - your actual sigma schedule, straight from whatever node builds it in your real graph. This is the whole reason to reach for this node over its siblings: feed it the literal object your sampler will use, not a reconstruction.
• context_opts (CONTEXT_OPTIONS, optional) - the plan you're checking. Leave unplugged to preview with no windowing applied, as a baseline.
• visual_width (INT, default 1440) - cosmetic width of the output chart.
• latents_length (INT, default 32) - how many frames the simulation assumes; you don't need an actual batch of that length sitting in the graph.

Output is a single IMAGE, same as the rest of the family - route it to Preview Image or Save Image.

How to install it

Through ComfyUI Manager: search AnimateDiff Evolved, confirm the author is Kosinkadink, install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved

then restart. This node doesn't need a motion module to run on its own, but the real generation you're validating against will.

Common issues & troubleshooting

You're building sigmas by hand just to use this node - don't. If your actual pipeline runs a plain KSampler or KSamplerAdvanced, constructing a SIGMAS object purely so you can preview is more work than the other two Visualize nodes require. Use Visualize Context Options (K.) for a standard sampler/scheduler run, or (K.Adv.) for a start/end step range - both take those settings directly and skip the sigma plumbing.

The preview looks fine but doesn't reflect your real run. This node is only as accurate as the sigmas you feed it. If you tap it off a branch of your graph and then later change the node upstream that builds those sigmas, the preview goes stale silently - there's no live link back to your real sampler.

Windows are only half the picture. As with the other two Visualize nodes, this checks the scheduling math, not the actual generated content. A well-formed window chart tells you your Context Options plan is internally sound; it won't catch a seam that comes from your motion module, ControlNet, or prompt schedule instead of the windowing itself.

You just want a sanity check, not the exact sigma trace. If getting your real sigmas plumbed to this node is more hassle than it's worth for a rough check, the "K." or "K.Adv." siblings will get you 90% of the same answer for a fraction of the wiring, since most custom sigma schedules still roughly track a recognizable sampler/scheduler shape.