NSFW-gen-v2 - AI Language Models Tool

Overview

NSFW-gen-v2 is a text-to-image model published by UnfilteredAI that is explicitly designed to produce uncensored, adult (NSFW) imagery from textual prompts. The model is hosted on Hugging Face and is presented as a StableDiffusionXL-based pipeline that accepts standard image-generation parameters (guidance scale, negative prompt, inference steps, size) while intentionally omitting safety filters so it can generate unmoderated outputs. ([huggingface.co](https://huggingface.co/UnfilteredAI/NSFW-gen-v2)) Technically targeted at creators, researchers, and moderation testers who need unrestricted generation, NSFW-gen-v2 is reported to run in FP16 for improved inference efficiency and includes prompt-friendly features such as a “3d”/“3d style” capability for more realistic rendering. The model card and recent repository updates note a ~3.47 billion parameter scale and describe the project as a text-to-image StableDiffusionXLPipeline derivative built on a PixelGen-style base. Users are repeatedly advised to employ the model only where explicit content is lawful and ethically permitted. ([huggingface.co](https://huggingface.co/UnfilteredAI/NSFW-gen-v2/commit/80d2ebda899c4f9443aeae617aa0369dbe76bba2?utm_source=openai))

Model Statistics

• Downloads: 5,832
• Likes: 494
• Pipeline: text-to-image
• Parameters: 3.5B

License: other

Model Details

Architecture & base: NSFW-gen-v2 is provided as a StableDiffusionXLPipeline (text-to-image) artifact on Hugging Face and is reported to use a PixelGen-derived base (OEvortex/PixelGen). The model files use safetensors and are optimized to run in FP16 precision for better memory/performance tradeoffs. ([huggingface.co](https://huggingface.co/UnfilteredAI/NSFW-gen-v2)) Size & capabilities: A repository update and model README indicate a parameter count of approximately 3.47 billion, placing it in the mid‑sized large-model bracket for image-generation fine-tunes. The project exposes standard diffusion controls (guidance_scale, negative_prompt, num_inference_steps, width, height, scheduler, seed) for fine-grained output steering and includes tags and examples showing explicit/uncensored generation as well as optional 3D-style rendering triggered via prompt engineering (e.g., include “3d” or “3d style”). ([huggingface.co](https://huggingface.co/UnfilteredAI/NSFW-gen-v2/commit/80d2ebda899c4f9443aeae617aa0369dbe76bba2?utm_source=openai)) Deployment & usage: The model can be run locally with diffusers-compatible pipelines or invoked via hosted inference endpoints (Hugging Face Inference or third‑party deployments such as Azure where available). The model card marks the repo as “Not-For-All-Audiences” and lists its license as “other,” so integrators should review licensing and platform terms before production use. File format, pipeline type, and language tags (English, Portuguese, Thai) are available on the Hugging Face page. ([huggingface.co](https://huggingface.co/UnfilteredAI/NSFW-gen-v2))

Key Features

• Uncensored, explicit-capable text-to-image generation with no built-in safety filter.
• StableDiffusionXL pipeline compatibility for high-quality image rendering.
• FP16-optimized model files for improved inference performance and lower memory use.
• 3D-style rendering triggerable via prompts (e.g., include “3d” or “3d style”).
• Multilingual prompt support noted for English, Portuguese, and Thai.

Example Usage

Example (python):

from diffusers import StableDiffusionXLPipeline
import torch

# Load the model (requires appropriate hardware and permissions)
model_id = "UnfilteredAI/NSFW-gen-v2"
pipe = StableDiffusionXLPipeline.from_pretrained(model_id, torch_dtype=torch.float16)
pipe.to("cuda")

prompt = "A cinematic portrait of a fantasy creature in soft natural light, highly detailed"
# Use negative_prompt to avoid unwanted artifacts; adjust guidance_scale and steps for quality/speed
result = pipe(
    prompt=prompt,
    negative_prompt="lowres, deformed, watermark, text",
    guidance_scale=7.5,
    num_inference_steps=30,
    width=1024,
    height=1024,
)

image = result.images[0]
image.save("output.png")

# Note: NSFW-gen-v2 is capable of uncensored outputs. Use responsibly and comply with laws and platform terms.

Benchmarks

Hugging Face likes: 494 (Source: https://huggingface.co/UnfilteredAI/NSFW-gen-v2)

Approximate parameters: 3.47B (Source: https://huggingface.co/UnfilteredAI/NSFW-gen-v2/commit/80d2ebda899c4f9443aeae617aa0369dbe76bba2)

Pipeline / task: StableDiffusionXLPipeline (text-to-image) (Source: https://huggingface.co/UnfilteredAI/NSFW-gen-v2)

Last updated (model page / listing): May 2025 (listing) (Source: https://ai.azure.com/catalog/models/unfilteredai-nsfw-gen-v2)

Key Information

• Category: Language Models
• Type: AI Language Models Tool