Shima Image Flip

Shima Image Flip is a mirror node: flips an image horizontally or vertically. That's the whole job, and it's the rare node where "the whole job" is genuinely useful because flipping is such a common composition move - fixing a subject facing the wrong way, doubling a batch for mirrored variations, or making a set of images read as a coherent pair.

The mechanism is unglamorous and fast: the image tensor gets converted to a numpy array and np.flip runs along axis 1 for horizontal or axis 0 for vertical, then it's converted back to a tensor. It's deterministic, CPU-only, and has no models behind it.

What to set

• image - the required IMAGE input.
• flip_type - the enum that matters: horizontal or vertical.
• enabled - the slightly sneaky third input. It defaults to true, and when false the node passes the image straight through untouched. That's the whole reason it exists beyond the obvious: you can leave a flip wired into a workflow and kill it from a boolean, or drive it from a Breaker/panel switch, without deleting the node or re-wiring.

Output is IMAGE, same shape as the input.

Where it fits

The classic ComfyUI habit: generate one image, flip it, and send both the original and the mirrored copy into a grid or a batch - instantly doubling your candidate set, which matters when you're doing a character-consistency pass or building a contact sheet. The enabled input also makes it a natural A/B control: flip the composition, run both versions, compare, and decide with the comparison in front of you instead of guessing.

One honest note: because the implementation calls .squeeze() on the batch dimension before flipping, this node is really built for single images - feed it a multi-image batch and you may end up with unintended shape behavior. If you're flipping a whole batch, test it on one image first. It's also worth saying plainly that this is exactly the kind of thing ComfyUI's core or any utility pack already has; Shima Image Flip earns its slot when you're already living in the Shima island system and want the consistent look plus the enabled gate, not as a reason to install the pack on its own.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/KDB-USJP/shima_wf

then restart, or ComfyUI Manager → search "Shima". No models, no extra pip packages for this one.