Shima Instagram Filters

Shima Instagram Filters is exactly what it sounds like: a dropdown of 26 named photo filters - _1977, aden, clarendon, earlybird, inkwell, kelvin, lofi, nashville, valencia, walden, xpro2, and friends - applied to your generated image. You pick one, your render comes out the other side looking like it was shot on a phone in 2013. It's a nostalgia machine, and honestly? For a certain kind of output it's the fastest way to make a flat, sterile render feel like a photograph.

Under the hood it's the pilgram Python library, which reimplements the classic Instagram filter set as PIL color operations. The pack auto-installs pilgram the first time the node runs (a lazy pip install inside the code - more on that below). Each filter is a deterministic transform, so you can run the same image through a row of filters and lay them out side by side for a cheap style sweep.

The inputs

• images - required IMAGE. Batch-friendly, like the pack's other image nodes; each frame gets filtered independently.
• instagram_filter - the enum with 26 choices. There's no "None" in the list, so you're committing to one of them. If you want the original, skip the node.

Output is IMAGE, same shape as the input.

Where it fits

Best used as a post-processing color-grade before save, usually on already-decoded images - a subtle aden or lofi can unify a batch, and the more extreme ones (inkwell is a straight B&W, moon too) are one-click dramatic looks. It pairs well with the pack's Hue Rotation: shift the palette, then bake in a film look. For a polished project you'll usually want to tune a real grade instead, but for speed and for the specific "shot on film" vibe these presets nail, nothing in core ComfyUI comes close without you hand-building the same matrices.

Install & the caveat that matters

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

then restart, or ComfyUI Manager → search "Shima".

The real gotcha is the dependency behavior: pilgram is installed on demand via a subprocess pip call inside legacy.py when the node first executes. That means (a) your first run of this node can hang for a bit while pip does its thing, and (b) it's the kind of auto-install that deserves a raised eyebrow - custom nodes run code as your user, and this is exactly the pattern the ComfyUI ecosystem's security incidents revolved around. It's a well-known, legitimate package, so the risk is modest, but know it's happening and keep the pack updated. If you're on a locked-down or air-gapped box, pre-install pilgram yourself before running the node. Beyond that, the only real annoyance is that there's no "disable" option - you pick a filter or you don't use the node.