Nothing to pay yet
Commission a LoRA
You bring the images, I do the training. I read the request first and tell you honestly whether a LoRA is even the right tool before quoting anything.
What I will not train
- Real people. Living or dead, famous or not, with or without their consent. No exceptions, no matter what you are offering.
- Anything sexualising minors, or characters designed to read as minors.
- Hate imagery, or models built to harass a specific person or group.
This is not a negotiation and asking twice will not change it. Everything else, ask — the worst that happens is I explain why it will not work.
Styles, characters, creatures, outfits, concepts: all fine. A style is not something anyone owns, and training on one is most of what I do.
What you get
- The .safetensors file itself, yours to keep and use
- Trigger words that actually work, and the weight range to use them at
- Sample generations, so you can see what it does before you rely on it
- One round of retraining if it misses the mark, included
What a LoRA cannot do
It nudges a model towards your subject; it does not guarantee it. Fine detail — exact logos, text, tiny asymmetric markings — comes out inconsistent no matter how good the dataset is. If your subject depends on that kind of precision, I will say so before taking money rather than after.
5 of 5 slots open
Commission tiers are being set up. Come back soon.