Fetching the paper…

FastComposer: Tuning-Free Multi-Subject Image Generation with Localized Attention · Around