A LoRA is a small trained file that steers a base image model toward a specific style, character, or look, without retraining the whole model. Whether a platform supports custom LoRAs is the main technical difference between one you can shape and one you can only prompt.
The name is short for Low-Rank Adaptation. The useful part is the size: a LoRA is a fraction of a full model, so it is cheap to train and cheap to swap.
Why it matters when choosing a platform
| Platform type | What you get |
|---|---|
| No custom model support | The house style. Consistent, and a ceiling |
| Curated LoRA library | A menu of styles someone else chose |
| Custom LoRA upload | Your own style or character, repeatably |
This is also why the same prompt gives different results across platforms: they run different base models and different fine-tunes, and the base version is rarely disclosed and changes without notice.
Where it connects
Custom LoRAs are how a consistent character survives across hundreds of images, which is the whole problem in AI influencer work. For general generation see NSFW AI image generators.
Part of the full NSFW AI safety checklist. Related: how to test a platform before paying.
FAQ
Do I need to train my own?
Only if you need a specific look repeatably. For most use, a curated library is sufficient and considerably less work.
Why does my prompt work on one platform and not another?
Different base models, fine-tunes, and prompt preprocessing. The version is rarely disclosed, so behaviour can shift under you without warning.
Is a LoRA the same as a model?
No. It is an adapter that modifies a base model. It cannot generate anything on its own.