# Base models and compatibility (/docs/models-loras/base-models)



Every image and video model is built on a particular architecture, called its base model, such as SDXL, FLUX, or Wan. Knowing the base model matters most when you add LoRAs.

## Why base model matters [#why-base-model-matters]

A LoRA is trained for one base model. It only works on presets built on that same base model.

* A LoRA trained for SDXL works on SDXL presets.
* That same LoRA will not work on a FLUX or Wan preset.

When you add a LoRA, set its **Base Model** correctly. If you try to use a LoRA on a mismatched preset, SimpliGen warns you, because the result would be broken or ignored.

<Callout type="warn" title="Mismatched base model">
  If the LoRA and preset base models do not match, SimpliGen shows a yellow &#x2A;*"Built for X"** badge on the LoRA. You can override it with **Use anyway**, but the output will usually be poor or corrupted. Always match base models before generating.
</Callout>

## Finding the base model [#finding-the-base-model]

* For a **preset**, the pack describes which model it uses.
* For a **LoRA** you download, the source page on Civitai or Hugging Face states the base model it was trained for. Use that value when adding it. See [Using LoRAs](/docs/models-loras/using-loras).

## Picking the right preset [#picking-the-right-preset]

If you have a specific LoRA or look in mind, choose a preset whose base model matches it. If you just want a result, pick the preset that fits the style and let it handle the model. You do not need to think about base models unless you are mixing in LoRAs.