Anima NSFW LoRA adapter drops on HuggingFace for unrestricted local generation

GAi92 released anima-base-1-loras-nsfw, a LoRA adapter that removes content restrictions from CircleStone Labs' Anima text-to-image model. The adapter landed on HuggingFace on May 17, 2026, tagged "not-for-all-audiences" and licensed under the same terms as the underlying Anima base model.

Anima is CircleStone Labs' open-weight text-to-image foundation. LoRA adapters let practitioners fine-tune specific behaviors—here, lifting safety filters—without retraining the full checkpoint. The adapter runs locally via the HuggingFace Diffusers library or ComfyUI custom nodes, a standard pattern for uncensored image generation. Because Anima is open-weight and the LoRA operates client-side, there is no server-side safety enforcement; practitioners retain full control over generation.

Local deployment

The model card shows zero downloads and zero likes at launch, typical for a fresh upload. GAi92 has not published sample outputs or benchmark comparisons, so early adopters will need to test prompt adherence and output quality themselves. The "other" license designation mirrors Anima's base-model terms; practitioners should review those before production use.

LoRA adapters typically add negligible memory overhead to inference. Practitioners running Anima locally can inject the adapter directly into the pipeline using standard Diffusers or ComfyUI workflows, making integration straightforward for existing Anima deployments.