Some additional training and minimized the watermark (not perfect).
This model was created with the recipe below.
( SDXL-Turbo + ( OnlyForNsfw118 - SDXL-Turbo ) ) + ( Latent Consistency Model (LCM) LoRA: SDXL : 0.5 ) + ( sd_xl_offset_example-lora_1.0 : 0.5 )
The recommended settings are
Sampling method : DPM++ 2M SDE Exponential or DPM++ SDE Karras
Sampling steps : 8~
CFG Scale : 1.5
This model seems to generate no worse than the original.
This model is overfitted for NSFW contents. Anything else will not work well. Enjoy and thank you.
Sorry for not listing all merged models.
Some additional training and minimized the watermark.
This model was created with the recipe below.
( SDXL-Turbo + ( OnlyForNsfw118 - SDXL-Turbo ) ) + ( Latent Consistency Model (LCM) LoRA: SDXL : 0.5 ) + ( sd_xl_offset_example-lora_1.0 : 0.5 )
The recommended settings are
Sampling method : DPM++ 2M SDE Exponential or DPM++ SDE Karras
Sampling steps : 8~
CFG Scale : 1.5
This model seems to generate no worse than the original.