Qwen Image 2.1 moves and resizes objects in images natively, no LoRA needed

linoy_tsaban · x · 2026-09-28

Developer linoytsaban found that the Qwen Image 2.1 base model can seamlessly move and resize objects in images out of the box — no LoRA needed. She had planned to finetune a LoRA for this task to showcase the new diffusers LoRA training support for Qwen 2.1, but the base model already proved strong enough. Demo and training links included.

Original post →

More from Multimodal

Multimodal channel →