Need that special anime desk lamp but don't know where to start?

AnimEasy is intended for generating sharp, flat-colored, (slightly-retro) anime images of, well... just about anything. Benchmarked on nearly 300 deliberately vague prompts from "happy" to "beach" to "microwave", AnimEasy has been optimized to always try and generate an anime-style image with clean line-art no matter what the prompt - no VAE, embedding or LoRa needed.

For best results use:

• Sampler: DPM++ 2M Karras (although they're all pretty similar)

• Steps: 10-20 (complex or obscure scenes may need more)

• Clip skip: 2 (1 still works but seems to leave out more details)

To upscale images use:

• Sampler: R-ESRGAN 4x+ Anime6B (Latent doesn't seem to like anime faces)

• Denoise strength: 0.5 (higher changes the scene more but may more detail)

Recommended (but not required)

• ADetailer (A1111 plugin, fixes up faces - use face_yolov8 models for best effect)

• ControlNet (to animefy an existing image, use ineart_anime + control_v11p_sd15s2_lineart_anime)

Below are some examples (50 steps, hi-res on, prompt in caption)

(ok, ok, I'll stop with the memes now)

Built primarily on top of BetterThanNothing's Toonify and BigBeanBoiler's Flat-2D-Animerge and with some minor contributions from several others in order to mainly tweak the eyes and base colors (includes a touch of CyberAlchemist's Anime Lineart and Thicker Lines LoRas)

N.B. This checkpoint isn't filtered or restricted and there are some seemingly innocuous prompts that can result in images that are a bit more... NSFW than you'd normally expect (like "chocolate" with a seed of 1) - you were warned, happy diffusing 😁

Bonus: Without clicking on the little "metadata info" icon, see if you can guess what prompts were used in the banner images (I may have cheated on a couple and used ControlNet to give SD more of a hint)