NSFW-gen-v2 text-to-image model debuts on HuggingFace with 3D and multilingual support
mmaatteusz released NSFW-gen-v2, an open-weight unfiltered text-to-image model supporting English and Portuguese, now available on HuggingFace as safetensors weights.
NSFW-gen-v2 is an unfiltered text-to-image diffusion model from creator mmaatteusz, now available on HuggingFace. The model runs on the Diffusers pipeline and ships as safetensors weights, tagged for 3D-style generation and multilingual support in English and Portuguese.
The model carries the "not-for-all-audiences" and "unfilteredai" flags, operating without built-in content restrictions—standard for open-weight diffusion models that practitioners can run locally and fine-tune without server-side safety enforcement. The safetensors format enables fast loading and memory-mapped inference, making it compatible with ComfyUI, Automatic1111, and Forge workflows.
The 3D tag suggests the model may be fine-tuned on synthetic renders or stylized outputs. Multilingual support for English and Portuguese is relatively uncommon in the text-to-image space, where most open-weight models default to English-only CLIP text encoders. The HuggingFace listing omits parameter count, base architecture, training dataset size, and benchmark comparisons—typical for early uploads from individual creators. As of May 13, the checkpoint had logged zero downloads and zero likes on the hub.