Kernel/sd-nsfw - AI Image Tools Tool

Overview

Kernel/sd-nsfw is a Stable Diffusion v1-5 variant on Hugging Face fine-tuned to emphasize photo‑realistic imagery and explicit (NSFW) content for research and controlled use cases. The model is distributed on Hugging Face under the CreativeML OpenRAIL-M license and is intended for researchers and practitioners who need a drop‑in text‑to‑image model tuned toward realism in adult content scenarios. It can be used directly for inference with the Diffusers pipeline or further fine‑tuned on downstream datasets when permitted by the license. Because this release is a focused fine‑tuning of Stable Diffusion v1-5, it retains the standard latent diffusion architecture and tooling compatibility (Diffusers, PyTorch) while altering learned weights to produce more realistic adult imagery. The model card on Hugging Face indicates modest community adoption (tens to hundreds of downloads/likes) and explicitly frames the model for research rather than broad commercial deployment. Users should review the CreativeML OpenRAIL-M license and Hugging Face model page for usage restrictions and safety guidance before downloading or fine‑tuning.

Model Statistics

• Downloads: 477
• Likes: 59
• Pipeline: text-to-image

License: creativeml-openrail-m

Model Details

Kernel/sd-nsfw is a diffusion‑based text‑to‑image model derived from Stable Diffusion v1-5. Architecturally it follows the latent diffusion paradigm used by SD v1-5: a text encoder conditions a UNet denoiser in latent space through timestep-conditioned denoising. The model is packaged for the Diffusers text-to-image pipeline, so it can be loaded with StableDiffusionPipeline.from_pretrained and run on CPU or GPU with PyTorch. The model card lists the pipeline as text-to-image and the license as CreativeML OpenRAIL-M (see the Hugging Face page). Specific internal parameter counts and fine‑tuning hyperparameters are not published on the model card, so parameter size is listed as unknown. Kernel/sd-nsfw is explicitly described as fine‑tuned for photorealism and NSFW outputs; because of that focus, downstream usage should follow ethical guidelines and license restrictions. Source: https://huggingface.co/Kernel/sd-nsfw and Hugging Face Diffusers documentation (https://huggingface.co/docs/diffusers).

Key Features

• Fine‑tuned from Stable Diffusion v1-5 for increased photorealism
• Optimized for generating NSFW/adult content for research scenarios
• Compatible with Hugging Face Diffusers text-to-image pipelines
• Supports direct inference and further fine-tuning when permitted
• Distributed under the CreativeML OpenRAIL-M license

Example Usage

Example (python):

from diffusers import StableDiffusionPipeline
import torch

# Load the model (public model example). Replace with your token if model is gated.
model_id = "Kernel/sd-nsfw"
pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16)
pipe = pipe.to("cuda")  # or "cpu"

prompt = "A photorealistic portrait of a person, studio lighting, 35mm"
image = pipe(prompt, guidance_scale=7.5, num_inference_steps=30).images[0]

# Save output
image.save("output.png")

# NOTE: This model is tuned for NSFW imagery. Follow the model license and ethical guidelines before use.

Benchmarks

Hugging Face downloads: 477 (Source: https://huggingface.co/Kernel/sd-nsfw)

Hugging Face likes: 59 (Source: https://huggingface.co/Kernel/sd-nsfw)

Pipeline: text-to-image (Diffusers) (Source: https://huggingface.co/Kernel/sd-nsfw)

License: creativeml-openrail-m (Source: https://huggingface.co/Kernel/sd-nsfw)

Parameters: unknown (Source: https://huggingface.co/Kernel/sd-nsfw)

Key Information

• Category: Image Tools
• Type: AI Image Tools Tool