GetFloatByIndex

The float-flavored twin of GetIntByIndex, and it exists for the same structural reason. The scheduling node at the front of the ComfyUI-InfiniteTalk-MultiImage workflow returns per-image values as lists, including real_start_time_list - the start times, in seconds, that each image's segment actually uses. The workflow's for-loop can't index into a list by itself, so MakeBatchFromFloatList packs the list into a batch and this node reaches in and pulls one value out per iteration.

GetFloatByIndex takes a float_batch (again, either a tensor or a plain Python list - the code checks for .tolist() and falls back to treating it as a list) and an index (default 0, minimum 0), and returns a single value as a FLOAT. The bounds handling is deliberately gentle: index past the end clamps to the last element, index below zero clamps to the first, and an empty batch returns 0.0 instead of raising.

In the shipped workflow it appears inside the loop with the loop counter wired to index, so each iteration reads the current segment's start time and hands it to the generation chain. You'll rarely touch it directly - it's plumbing - but it's the node that makes "each image starts when I said it starts" actually happen.

Inputs: float_batch (FLOAT), index (INT). Output: value (FLOAT). Nothing else.

Install is the pack's standard one-liner:

cd ComfyUI/custom_nodes
git clone https://github.com/xuhongming251/ComfyUI-InfiniteTalk-MultiImage

or ComfyUI Manager → search ComfyUI-InfiniteTalk-MultiImage → Install → restart. The pack has no Python dependencies and downloads no models - all four of these utility nodes are pure graph glue.

One honest caveat, same as its integer sibling: the clamping means a bad index never announces itself. If your segments are all starting at the wrong times, don't debug this node - count the entries in the batch and check the index source. And remember the default index is 0, which for a multi-segment workflow is almost always not what you want if you wired it by hand; the loop needs to increment it.