Diffusion Docs

Appearance

LoRA Compatibility ​

LoRA compatibility depends on more than the filename. It depends on the model family, architecture, text encoder, training target, and sometimes the UI/backend.

Quick Take

If a LoRA fails, first verify compatibility with the base model. Prompt changes cannot fix a fundamentally mismatched adapter.

Compatibility Checklist ​

  • model family or architecture matches
  • expected base model is known
  • trigger words are documented
  • UI/backend supports the LoRA type
  • weight is in a reasonable range
  • no conflicting adapters are loaded

Symptoms Of A Mismatch ​

  • no visible effect
  • severe artifacts at normal weights
  • strange colors or textures
  • prompt ignored when the LoRA is enabled
  • errors when loading the file

What To Verify ​

Check the LoRA page or model card for:

  • base model family
  • recommended checkpoint, if any
  • trigger words
  • example prompt
  • recommended weight range
  • whether it targets style, subject, character, clothing, pose, or another concept
  • known incompatibilities

If this information is missing, test cautiously and save notes.

Architecture Mismatch ​

Architecture mismatch is more serious than prompt mismatch. A LoRA made for one architecture may fail to load or may produce meaningless changes in another. A UI allowing file selection does not guarantee semantic compatibility.

Text Encoder Mismatch ​

Some LoRAs rely on trigger words or learned associations through a specific text encoder. If the target model uses a different encoder, the trigger may not activate the intended concept.

Model Family Matters

Compatibility names are ecosystem-specific. Do not assume a LoRA made for one generation family will work with another simply because the UI lets you select it.