Image Style

Image Style is a nostalgia node. It applies the classic Instagram filter set - 1977, Clarendon, Valencia, X-Pro II, all 26 of them - to your generated image using the pilgram library, which reimplements those old looks with numpy math instead of a phone app. Select a filter from the style dropdown and every image in your batch comes out with that warm, faded, slightly-too-contrasty vibe.

What it's actually good for

Honestly? Mostly vibes. These filters are the same presets that made every 2012 travel photo look like a memory, and they're a fast way to give generations a consistent, film-ish finish without touching a LUT or a color-grade node. If you're building a batch of "same scene, different mood" variants - Clarendon for bright and punchy, Inkwell for black and white, Nashville for that pink-washed look - it's a one-dropdown solution.

The real power move is the All boolean. Flip it on and the node ignores style entirely, running every filter across every input image, and stacking the results into one big batch. Feed it a single image and you get a 26-image batch, one per filter - a contact sheet of every look at once. That's a genuinely useful way to pick a direction before committing to one, and it's the reason this node earns its keep beyond being a toy.

A couple of honest caveats. It's a per-image, CPU-side pixel operation - no model involved, no VRAM cost, and it's fast, but the results are exactly as subtle as the 2010s filters were (i.e., not very). And because it ships from the era of "just pip install it at import time," the node literally runs pip install pilgram as a subprocess if the library isn't found when the pack loads. That's the only heavy-ish dependency in the pack, and if pip fails mid-import you'll see a scary traceback on startup. If you install the pack's requirements.txt properly, you'll never hit it.

One quirk worth knowing

The dropdown lists 26 filters and the code tries to map each to a pilgram function. The mapping for "toaster" actually applies stinson - a copy-paste slip in the source that survives in both the single and All paths. So if you pick Toaster you'll get Stinson's look. It's a tiny thing, but if you're comparing filters side by side, it'll throw your eyes off.

What's on the node

• image - the IMAGE input (a batch works; every image in it gets filtered).
• style - the filter dropdown (26 choices).
• All - optional boolean (default off): apply every filter to every image and stack the output.
• IMAGE output - the filtered image(s).

Install

cd ComfyUI/custom_nodes
git clone https://github.com/r-vage/ComfyUI-RvTools_v2
# restart ComfyUI

Or via ComfyUI Manager, searching "ComfyUI-RvTools_v2". If the pack's dependencies aren't installed by your manager, you'll need pip install pilgram - it's the one non-standard library this node leans on. No model downloads.

The pack caveat

Like every node here, it lives in a pack the README flags as unmaintained, pointing to ComfyUI_Eclipse as the successor. The original RvTools repo was pulled from GitHub in early 2025, breaking workflows that referenced it. v2 installs cleanly and keeps those workflows alive; for new builds, Eclipse is where this author's tooling now lives.