My humble experiment with LoRa training. Works well with EpicPhotogasm checkpoint.Let me introduce Natalia Nemchinova the sexiest cheerleader on FIFA 2018 and erotic model, she also likes swinging, fetish parties etc. <h2 id="heading-1058">Usage</h2>The following settings are just recommendations that work well for meCheckpoint: EpicPhotoGasm version Last Unicorn <a target="_blank" rel="ugc" href="https://civarchive.com/models/132632/epicphotogasm?modelVersionId=223670"> (Link)</a>Sampler: DPM++ 2M SDE Karras Sampling steps: 25Resolution: 512x512Clip skip: 1VAE: Automatic<h3 id="heading-1059">Hires, fix (optional, but recommended)</h3>Hires steps: 20Denoising strength: 0.3Upscaler: R-ESRGAN 4x+ (or any other AI upscaler)Upscale by: 2 (haven't tested with higher values, because of VRAM limitations)<h2 id="heading-1060">Versions</h2>v1.0 is SFW for face only generationv2.0+ is NSFW

Natalia Nemchinova aka Danica (erotic model)

Now the model is able to generate full body, NSFW content added as well

Files with SHA256: b96ec2f1c1fffd4cfd976f26ae81519c694e05f24de26c59dc39caa8b74fc52a