Turbo version should be used at CFG scale 2 and with around 4-8 sampling steps. Should work with DPM++ SDE Karras (or Normal). Comparison of other samplers: https://civitai.com/posts/951781
Can be used for highres fix and tiled upscaling. Please check the examples.
If you train on this, make sure to use DPM++ SDE.
What's new in 2.1 Fixes a skin issue that sometimes occurred in v1 and v2. I used targeted adversarial learning for it. The problem didn't affect SFW versions of the model, so I won't update that one. It should have little to no effect on most images compared to v2, except other minor improvements on anime stuff apparently (which was totally collateral).
Turbo version should be used at CFG scale 2 and with around 4-8 sampling steps. Should work with DPM++ SDE Karras (or Normal). Comparison of other samplers: https://civitai.com/posts/951781
Can be used for highres fix and tiled upscaling. Please check the examples.
If you train on this, make sure to use DPM++ SDE.
What's new in 2.1
Fixes a skin issue that sometimes occurred in v1 and v2. I used targeted adversarial learning for it.
The problem didn't affect SFW versions of the model, so I won't update that one.
It should have little to no effect on most images compared to v2, except other minor improvements on anime stuff apparently (which was totally collateral).