Release of Image-to-LoRA (i2L) V2 by DiffSynth-Studio

🎨 Release of Image-to-LoRA (i2L) V2 by DiffSynth-Studio

The company has released a technology that allows for the creation of LoRA models based on one or more reference images. The new version is based on the Z-Image architecture (a successor to Qwen-Image-i2L) and significantly improves style preservation accuracy. The tool supports SD1.5 and SDXL.

🌍 The technology simplifies the fine-tuning process, replacing lengthy training with instantaneous weight generation from a visual sample, which lowers the barrier to entry for creating stylistic models.

👤 It is no longer necessary to collect hundreds of images and wait hours for training to transfer a style into Stable Diffusion. Simply provide a reference to obtain a ready-to-use LoRA file.

Source 1: https://www.modelscope.cn/models/DiffSynth-Studio/Z-Image-i2L/summary Source 2: https://github.com/modelscope/DiffSynth-Studio