Pick Image From Batches (FS)

The README's own description is the simplest way to put it: "this node will pick one or a couple images from a list of batches." It's the third member of the pack's pick-an-image family, alongside "Pick Image From Batch" (a single stacked batch) and "Pick Image From List" (a Python list of individual images) - this one is for when you've got several batches at once and want to pull a selection across them rather than combining them first.

Why it exists

If you're running multiple generations that each produce their own batch - say, a few different seeds or a few different samplers, each outputting its own IMAGE batch - and you want to sample across all of them without manually merging batches into one tensor first, this is the node built for that. It saves you a combine-then-pick step.

How it works

Same selection mechanism as its siblings in this pack: seed plus start_index control whether the pick is random or fixed, and select_amount controls how many images come out. start_index at its default of -1 means a random, seeded pick; set it to 0 or higher to start deterministically at that position instead.

The inputs and outputs that matter

• seed (INT, default 0) - drives the random selection.
• start_index (INT, default -1, range -1 to 32) - -1 for random, 0+ to fix the starting position.
• select_amount (INT, default 1, range 1 to 32) - how many images to pull.

Output: IMAGE - your selected image or small set. Note this node's fixed settings are just those three: the batches themselves connect in through the node's image input slots on the canvas rather than a single named field, which is why they don't show up as a separate named parameter here - connect your batch sources to it directly and the three controls above decide what gets picked out of the combined pool.

Installing it

ComfyUI Manager: search "Comfy Fit Size" or "fitsize." Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/bronkula/comfyui-fitsize.git

Restart ComfyUI. No model downloads, nothing heavy - small utility pack. Lives under Fitsize → Image.

Common issues

This is the least-used node in the pack by a wide margin, and there's essentially no community discussion of it to draw on - so treat the following as general guidance rather than reported gotchas. The same -1 = random (not "last") convention applies here as it does across the whole pick-image family, and both index fields cap at 32. If you're not seeing the inputs you expect on the node, check that you've actually connected batch sources to it - unlike "Pick Image From Batch," there's no single required image input doing the heavy lifting, so an unconnected node here won't have anything to pick from.