Noct Q Anime
Uncensored anime on Qwen Image 2.1: soft light, clean color, and full nudity when the scene calls for it.
What it does well
- Anime illustration, from close portraits to wide background paintings.
- Adult characters dressed, in lingerie or nude, and explicit scenes.
- Long, detailed prompts followed closely: two characters, poses and camera angles as written.
How to prompt
Start with "An anime illustration of…" and describe the scene in full sentences. Without the word "anime" you get a photo, as with Noct Q. No trigger word.
Which file
int8 for 8 to 12 GB cards, int4 for smaller ones; fp8, bf16 or fp16 otherwise.
Settings
25 steps, euler, simple, CFG 3 with the negative prompt from the included workflow. Same text encoder and VAE as Noct Q: text encoder, VAE.
Built with Qwen.
中文：基于 Qwen Image 2.1 的无审查动漫模型。提示词以 "An anime illustration of…" 开头，用完整句子描述画面。推荐 25 步、CFG 3，使用附带工作流中的负面提示词。
Description
First release: uncensored anime on Qwen Image 2.1. Start prompts with "An anime illustration of…". Files: int8, int4, fp8, bf16 and fp16, plus a zip with the licence and a ComfyUI workflow.
FAQ
Comments (9)
Nice. Early Qwen Image 2.1 has been lacking a bit in the anime Loras. Can't wait to sink my teeth into it.
very nice, al tho can good anime charters, but does not have any data image from anime shows yet, but this just the beginning hope you can improve this model even better, look forward for version 2
one for the edit model?
This one is not really meant for image edit, right? Also, it seems to be hit & miss when asking for transparency; tends top generate the checkered background most of the time.
can this model use for edit?
This model might be a hidden gem...
It can do edit. That's the thing of qwen image 2.1. Though I'm not sure if this one was specifically for editing. Tried cloth removal with furry and human characters, both work, though body modifications seem slightly off like with the base qwen 2.1. Probably could use further training, but still very good!
Too bad this one also doesn't seem knowledgeable of male anatomy like the rest...
Hope you add Turbo version
Are we supposed to use the text encoder and VAE listed there, or is it fine to just use the official text encoder and VAE from Qwen Image?