Why doesn't my output match my reference?

Different models interpret references differently - some prioritize identity, others composition or style. Identity drift is a known behavior of current-generation AI models, not a PromptWise setting.

To improve consistency:

• use a Seed instead of raw reference images
• use models known for identity preservation
• generate multiple variations and select the best
• keep prompts specific about which reference controls what (via @ mentions)

Updated on: 07/07/2026