ComfyUI NSFW Detector

A ComfyUI custom node that detects NSFW content using Freepik/nsfw_image_detector (EVA-based, ~86MB) and saves images with detection results in the filename.

Node: NSFW Detector + Save Image

Drop-in replacement for SaveImage with built-in NSFW detection.

Inputs

Name	Type	Description
images	IMAGE	Input images
filename_prefix	STRING	Filename prefix (default: ComfyUI)
nsfw_threshold	low / medium / high	Detection threshold (default: medium). See Threshold levels

Outputs

Name	Type	Description
images	IMAGE	Passthrough (for chaining)
nsfw_level	STRING	Detected level: neutral / low / medium / high
nsfw_score	FLOAT	Confidence score of the detected level

Threshold levels

The model classifies images into 4 levels: neutral, low, medium, high. The nsfw_threshold parameter controls the minimum level that triggers a flag — images at or above the threshold are flagged.

Threshold	Flagged levels	Passed levels	Strictness
low	low, medium, high	neutral only	Strictest
medium	medium, high	neutral, low	Balanced
high	high only	neutral, low, medium	Most lenient

Filename convention

• Normal: {prefix}_00001_.png
• Flagged: {prefix}_flagged_{level}_00001_.png

Installation

Clone into ComfyUI's custom_nodes directory:

cd ComfyUI/custom_nodes
git clone https://github.com/newraina/comfyui-nsfw-detector.git
pip install -r comfyui-nsfw-detector/requirements.txt

Pre-download model

The model is downloaded from HuggingFace on first use by default. To pre-download, place model files in ComfyUI/models/nsfw_detector/:

models/nsfw_detector/
  config.json
  model.safetensors

If you're using comfyui-api, add to manifest.yml:

models:
  before_start:
    - url: https://huggingface.co/Freepik/nsfw_image_detector/resolve/main/config.json
      local_path: models/nsfw_detector/config.json
    - url: https://huggingface.co/Freepik/nsfw_image_detector/resolve/main/model.safetensors
      local_path: models/nsfw_detector/model.safetensors