(Note: The Lora weight on the left side of the example image is 0)
Training resolution : [2560 × 1472],train base on Anima base1.0
Effect: Provide clearer results for images with resolutions of 2K and below, slightly enhancing image aesthetics and lighting effects,
Beside, it has a relatively small impact on the style and composition of the image
Improve image stability when directly generating images with a resolution of 1536 or higher(2 million pixels)
Disadvantage: Sometimes it can cause color cast in the image
Suggested weight: [0,2] In general, using 1 weight has a good effect.
Description
FAQ
Comments (8)
滑块的神
咋用?当普通lora用还是当放大模型用?
Tested this at 1216x832 (so not 2k) with a few random LoRAs when the hands broke and it stabilized the image pretty well without breaking the style/composition. Nice work here!



