Default Iteration Options ππ π
The plain baseline behind FreeInit
- ITERATION_OPTS
The README's whole mention of this feature family is one line: "FreeInit and FreeNoise support (FreeInit is under iteration opts...)." FreeInit - a real academic technique that resamples a clip a few times, reusing the low-frequency structure of each pass to fix flicker - lives on a sibling node, ADE_IterationOptsFreeInit. This node is the other member of that same family: the plain baseline. Wire this in and you get straightforward multi-pass sampling, without FreeInit's frequency-domain noise mixing.
What it actually does
Both nodes build an ITERATION_OPTS object that plugs into the iteration_opts slot on Sample Settings (ADE_AnimateDiffSamplingSettings), which in turn feeds a loader or Use Evolved Sampling. This node's job is running your sampling pass iterations times over, offsetting the seed and batch position between passes so each repeat isn't just an identical copy of the last one - rather than FreeInit's more involved noise-blending across iterations.
That makes it the node to reach for when you want the cost of resampling - a chance for the model to land on a cleaner result across a couple of tries - without paying for FreeInit's more targeted (and more expensive) flicker-fixing machinery. If you don't need either, leave iterations at 1 and this whole family is a no-op you can ignore.
The inputs and outputs that matter
Required: iterations (default 1, minimum 1) - how many full sampling passes to run. This is a direct cost multiplier: setting it to 2 roughly doubles your render time for the clip, same as it does on the FreeInit variant.
Optional: iter_batch_offset (default 0) and iter_seed_offset (default 0) - shift the batch position and seed between each iteration, so repeat passes aren't sampling from identical starting noise. Leave both at 0 if you specifically want every iteration to start from the exact same conditions.
Output is a single ITERATION_OPTS, and it only fits Sample Settings' iteration_opts input.
How to install it
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. No model download tied to this node - it's pure sampling configuration on top of whatever motion module you've already got loaded elsewhere in the graph.
Common issues & troubleshooting
Set iterations above 1 and rendering takes noticeably longer. That's expected and unavoidable - each iteration is close to a full additional sampling pass. If the render-time cost isn't earning its keep, drop back to iterations: 1 and look elsewhere for whatever quality problem you were trying to solve.
Wired this in expecting FreeInit's flicker-reduction and didn't get it. This is the plain variant, not FreeInit. If temporal flicker or boiling detail is specifically what you're chasing, you want ADE_IterationOptsFreeInit instead - it does the actual frequency-domain noise blending that targets that problem. This node just repeats the sample.
Not seeing any difference at all with iterations above 1. Confirm the ITERATION_OPTS output actually reaches Sample Settings' iteration_opts slot, and that Sample Settings itself feeds your loader or Use Evolved Sampling. None of these connector nodes error loudly when left unplugged, so a broken chain here fails silently rather than obviously.
Worth being upfront: iteration opts as a whole is a deep-cut part of the pack. Most people troubleshooting AnimateDiff reach for a better context-window setup or a different motion module before they reach for repeated sampling passes - this is a tool for a specific, already-diagnosed problem, not a general first move.
Inputs (3)
| Name | Type | Default | Description |
|---|---|---|---|
| iterations | INT | 1 | β |
| iter_batch_offsetopt | INT | 00β9007199254740991 | β |
| iter_seed_offsetopt | INT | 0-9007199254740991β9007199254740991 | β |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| ITERATION_OPTS | ITERATION_OPTS | β |