RopicsXL — Rocafort × Prédondo Western Comic Style LoRA

Trigger: ropicsxl

Recommended weight: 1 (can push to 2 if you like it punchier)

CFG: 2–7.5 (I use 6)

This LoRA is a fusion of Kenneth Rocafort + Predondo, tuned to hit that clean western-comic vibe — sharp lines, tight anatomy, punchy poses, expressive faces, shiny highlights, all that good stuff. Buzz, grit, stylized flare… it’s all welcome. That’s the fun of comic art.

This is NOT a NSFW LoRA by design. NSFW works fine, but the LoRA doesn’t inject porn style — it defers to your base checkpoint for that.

Good for pinups, superhero panels, portraits, covers, and anything with comic energy.

Characters it knows well (tagged in dataset)

These were actually tagged, so they respond to name prompts:

Starfire

Raven

Supergirl

Superman

Madame Mirage

All characters come mostly from the New 52 era, so you might see some New 52 outfit influences pop up.

Supergirl may sometimes show her Red Lantern variant, because yeah… some of those were in the dataset. Oops, lol.

Merging / Permissions (Read This)

I turned on “merge allowed” because… honestly? LoRAs were made to be merged.

But YOU take the risk. If you merge it into a franken-model and it explodes into spaghetti anatomy, that’s on you, not me.

I’m not going to pretend this LoRA is some untouchable masterpiece — it’s solid as hell, but it’s not perfect.

If you wanna merge it, go wild. If it melts… you were warned. 🤣

Checkpoint Compatibility

Works with any IllustriousXL-based checkpoint.

If it’s built on IllustriousXL, this LoRA runs clean.

Training Specs (for the nerds)

Images: 998

Epochs: 16

Repeats: 1

UNet LR: 0.0003

Text Encoder: 0.0001

Scheduler: Cosine with Restarts (3)

Dims: 64 / 32

Noise Offset: 0.05

Final Notes

This was made for fun, to experiment, and to see how well two western comic styles could fuse.

Turns out… they fuse stupidly well.

Enjoy it, break it, merge it, abuse it—whatever.

That’s the whole point of AI art: mess with styles and create cool shit.