Noctaluna · Uncensored image model

Noct Q: Qwen Image 2.1 Uncensored Realism (V4)

Noct Q is a community text-to-image model from Noctaluna, built as a merge of Qwen/Qwen-Image-2.1 with modified transformer weights. It is tuned to render nudes and explicit adult scenes without a LoRA, and also handles readable text on signs and labels, anime and painted illustration, and image editing with the same files in ComfyUI's Qwen Image 2.1 edit template. The card says V4 renders explicit scenes with a partner as asked three times as often as V3, but gives no benchmark or method behind that figure. The V4 int8 convrot file is 7.3 GB and is aimed at 8 to 12 GB cards, with ComfyUI 0.37 or newer and core nodes only. Recommended settings are 25 steps, euler, simple, cfg 3 with a negative prompt, and 1024x1536 portrait; cfg 1 runs about twice as fast but ignores the negative prompt. Prompts should be full-sentence captions rather than tag lists, and no trigger word is needed. It was created on 2026-09-28 and has 653 likes. It is released under the Qwen Research License, for non-commercial use only; int4, fp8, bf16 and fp16 files are on Civitai. The model is tagged not-for-all-audiences.