- Anime model is now available! For better results use danbooru tags, but it also works with standard prompts. - Adjusted UI a little bit. Less accent colors, limited height of the init image, smaller buttons - Added scroll up button
- Added slider for # of images to generate - Added ddos, bots protection - Optimized the website, it'll load faster and work better - Bug fixes - Disabled the NSFW filter - Updated to SD 1.5
Added steps & sampler selection. k_heun, k_dpm_2 and k_dpm_2_a are 2x slower, but converge faster, so they are capped at 30 steps, while other samplers are capped at 50 steps.