Content Moderation
Review user-generated media instantly before it reaches your platform. Guard automates your content moderation through a powerful API. Discover how to scan images, what you can detect, and how to automate your response.
Problem
Unfiltered Media Problem
User-generated content brings unpredictable risks. No matter what rules you set, inappropriate, violent, or synthetic media will inevitably be submitted to your servers. This constant influx is the biggest hurdle for any online space where users connect, trade, or share opinions.
Marketplaces
Overwhelmed by synthetic images and scam listings
Social Media Platforms
Infiltrated by graphic content and unverified uploads
Review Sites
Bombarded with deceptive media and coordinated fake reviews
Damages
One Step Behind
Manual review is simply too slow for a growing platform. By the time your team catches offensive or misleading media, it has usually already reached your users. Relying on human moderation alone guarantees that toxic content will slip through, triggering a chain reaction of negative consequences for your business.
Damaged Reputation
Destroys your brand image with just a single toxic upload
Mass User Churn
Drives away users who no longer feel safe on your platform
Lost Revenue
Scares off legitimate buyers, creators, and valuable advertisers
User Story 1
Automated Media Scanning
Send images and videos to the Guard API for automatic, server-side analysis. You can configure it to detect a variety of critical concepts — for instance, instantly identifying and blocking AI-generated content.
$ python scan_media.py --media uploaded_image.jpg
media: uploaded_image.jpg
space: My Demo Space
results:
- AI-Generated
score: 0.984637
description: The texture of the skin and the lighting on the background do not match any capture device. Several regions carry the repeating patterns typical of a diffusion model.
User Story 2
Multi-Concept Detection
Need to filter for more than just AI? Combine our core tasks to scan media for several conditions simultaneously. Just configure your custom spaces to apply the exact detection models you need, checking every upload against all your rules in a single pass.
$ python scan_media.py --media uploaded_image.jpg --space b920d254-f254-40e4-8d63-ad59f31c622f
media: uploaded_image.jpg
space: My Inbound Space
results:
- AI-Generated
score: 0.964812
description: Repeating diffusion patterns run through the background and the skin texture.
- Violent
score: 0.912774
description: A physical assault with visible injuries dominates the foreground.
- Explicit
score: 0.873109
description: Exposed intimate areas are visible in the centre of the frame.
User Story 3
Set Your Own Moderation Rules
How you act on the analysis results is entirely up to you. For example, you can set a high confidence threshold for careful, conservative filtering. This ensures that when content gets flagged, it is almost certainly a true match, allowing you to prioritize extreme accuracy while leaving borderline cases alone.
$ python moderate_media.py --media-dir media --block-ai 0.9 --review-ai 0.8 --block-violent 0.6 --review-violent 0.5
65 media blocked
9 media need review