Taco Img2Img Animated Processor

Taco Img2Img Animated Processor is the in-graph version of the pack's img2img batching trick. Where Taco Img2Img Animated Loader reads a still from disk and repeats it, this node takes an IMAGE tensor from anywhere in your workflow - the VAE-decoded output of a KSampler, say - and repeats it frames times into a batch for further processing. No file picker, no input folder, just "here's an image, now there are eight of them."

It sits under Taco_Nodes/Gifs, and it's the newest node in the pack. The README's example workflow shows the intended shape: generate an image with KSampler, decode it with a VAE, feed the result into this processor, then run the batch through more processing (img2img, batch-level effects) and finally into Taco Gif Maker to recombine.

Inputs and output

• image (IMAGE) - any image tensor from the graph. Because it's a tensor, you can batch after generation rather than before, which is the point: start with a finished render and build the animation from it.

• frames (INT, default 8, 1–1000) - how many copies of the input to make.

• IMAGE (output) - the input repeated frames times along the batch axis.

How it works

The mechanism is genuinely a few lines: it takes the incoming tensor and concatenates frames copies of it down the batch dimension (torch.cat over dim 0). If your input is already a batch of N images, you get N × frames out - which is usually not what you want, so keep the input a single image unless you're deliberately building a bigger batch.

One honest warning about the name: "Img2Img" is descriptive of the workflow this node was made for, not something the node itself does. It does zero image-to-image processing. It's a batching utility, full stop. If you were expecting a sampler or interpolation inside, that's not here.

How this fits an animation workflow

The pattern this enables is the classic per-frame img2img loop: generate one good frame, batch it N times, then run the batch through img2img with per-frame seed or denoise variation so each copy drifts slightly, then recombine into a GIF. It's the "spice up a still" trick done from a tensor rather than a file. Remember that batching alone produces identical frames - the motion only appears if something downstream varies the result per batch position.

Installing it

It ships in ComfyUI-TacoNodes. Install via ComfyUI Manager by searching "ComfyUI-TacoNodes", then restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/YOUR-WORST-TACO/ComfyUI-TacoNodes.git

Restart and it appears under Taco_Nodes/Gifs. There's no requirements.txt and no model download - the pack relies only on PIL, numpy and torch, all bundled with ComfyUI. It's a tiny personal pack (Apache 2.0, single maintainer) with almost no community footprint, so don't expect forum help; the GitHub repo is the only real documentation. It's a thin utility, but if you're building the "animate a generated image" family of workflows in this pack, it's the piece that lets you start from inside the graph instead of from a saved file.