A LoRA is a lightweight AI image add-on that fine-tunes a base model for specific styles or characters. Learn how LoRAs work and use them on CRAISEE.
Frequently asked questions
What does LoRA stand for in AI?
LoRA stands for Low-Rank Adaptation. It is a technique introduced in the 2021 paper 'LoRA: Low-Rank Adaptation of Large Language Models' by Hu et al. In AI image generation, it refers to a small add-on file that modifies a subset of a base model's weights to produce specific styles, characters, or concepts.
How big is a LoRA file?
LoRA files are typically between 1 MB and 200 MB, making them far smaller than full AI models, which can be several gigabytes. This compact size means LoRAs are easy to share, download, and swap without significant storage or computational overhead.
What is the difference between a LoRA and a base model?
A base model like Stable Diffusion XL or Flux is the core AI that generates images. A LoRA is an add-on that layers on top of the base model to redirect its output toward a specific style or character. A LoRA cannot function independently — it always requires a compatible base model to run.
Can you use multiple LoRAs at the same time?
Yes, multiple LoRAs can be stacked in a single image generation. For example, you can combine a lighting-style LoRA at a weight of 0.6 with a character-concept LoRA at 0.5, and both will simultaneously influence the output. Balancing weights carefully prevents conflicting effects.
What is the LoRA weight slider and what does it do?
The LoRA weight slider controls how strongly a LoRA influences the generated image. A value of 0.0 means the LoRA has no effect, while 1.0 applies its full influence. Values between these extremes blend the LoRA's style with the base model's default output, giving fine-grained creative control.
Do I need to download LoRA files to use them on CRAISEE?
No. On CRAISEE, LoRAs are available directly in the Model & Add-ons panel during image creation. You do not need to download files or manage local directories. Every LoRA in the CRAISEE library can be attached to your generation workflow with a single click.
What base models are compatible with LoRAs?
LoRAs are designed for specific base models and must match to work correctly. Common compatible base models include Stable Diffusion XL (SDXL) and Flux. On CRAISEE, the platform indicates which LoRAs are compatible with each supported base model, preventing mismatched pairings.
What can LoRAs be used for in AI image generation?
LoRAs can be used to generate specific art styles, recreate characters, apply consistent lighting or color palettes, and introduce niche visual concepts that prompting alone cannot reliably produce. They are widely used for portrait styles, anime aesthetics, product visualization, and brand-consistent image creation.
