Context OptionsโStandard Static ๐ญ๐ ๐
Fixed chunks instead of a sliding window
- prev_context
- view_opts
- CONTEXT_OPTS
Same job as ADE_StandardUniformContextOptions - get AnimateDiff past its native 16-frame window - but a simpler mechanism. Uniform context slides an overlapping window across your whole sequence with a configurable stride; Static context chunks the sequence into fixed blocks instead. You can tell the difference in the schema alone: this node drops context_stride entirely, because a static schedule doesn't need an interleaving offset the way a sliding one does.
Practically, most public AnimateDiff workflows default to Uniform, because that's the schedule behind the original "unlimited context length" feature that made the pack famous. Static is worth reaching for when you specifically want more predictable, blockier chunking behavior - fewer moving parts, less blend logic between windows - rather than the smoother but more compute-heavy overlap dance Uniform does.
How it works
Your frame sequence gets divided into context_length-frame chunks, with context_overlap frames shared between adjacent chunks for blending at the seams. Without a stride parameter, there's no interleaved offset pattern - it's a more literal chunk-and-blend than Uniform's continuously sliding window. VRAM cost still scales with context_length, not your total clip length, so the core benefit (arbitrary-length generation on fixed VRAM) carries over unchanged.
The inputs and outputs that matter
Required: context_length (default 16, matched to your motion module's native window) and context_overlap (default 4, controls how much adjacent chunks share for blending). That's it for required fields - noticeably fewer than the Uniform variants.
Optional, same shape as the rest of the Context Options family: fuse_method for blend behavior at chunk boundaries, start_percent and guarantee_steps for scheduling this context config to specific phases of sampling, prev_context for chaining multiple context configs, and view_opts for a nested inner sliding window on top. Output is CONTEXT_OPTS, feeding a loader's or Use Evolved Sampling's context_options input.
How to install it
Standard for the pack - ComfyUI Manager, search AnimateDiff Evolved by Kosinkadink, or:
cd ComfyUI/custom_nodes && git clone https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved
No model download tied to this node specifically - it configures sampling behavior on top of whatever motion module is loaded elsewhere in your graph.
Common issues & troubleshooting
Visible chunk boundaries in the output. Same fix as Uniform: raise context_overlap so adjacent chunks share more frames for the blend to work with, or lower denoise slightly on the affected pass.
Not sure whether to pick this over Uniform. If you don't have a specific reason to want Static's blockier behavior, default to ADE_StandardUniformContextOptions - it's the schedule most community workflows and tutorials build around, so you'll find more example workflows and more troubleshooting discussion referencing it if something goes wrong.
Set context_length far above 16 hoping for smoother long clips. Doesn't help - the module's native training window is still 16 frames, and pushing this number up tends to hurt coherence rather than improve it, on either schedule type.
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| context_length | INT | 161โ128 | โ |
| context_overlap | INT | 40โ128 | โ |
| fuse_methodopt | COMBO | 10 options: pyramid, relative, flat, overlap-linear, ๐ฌdelayed reverse sawtooth, ๐ฌpyramid-sigma, +4 | |
| use_on_equal_lengthopt | BOOLEAN | false | โ |
| start_percentopt | FLOAT | 0.0000โ1 | โ |
| guarantee_stepsopt | INT | 10โ9007199254740991 | โ |
| prev_contextopt | CONTEXT_OPTIONS | โ | |
| view_optsopt | VIEW_OPTS | โ |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| CONTEXT_OPTS | CONTEXT_OPTIONS | โ |