Image moderation built for user-generated content.

Analyze images to detect sensitive, harmful and inappropriate visual content, including AI-generated content, QR codes and CSAM. Bodyguard combines detailed image classifications with text moderation through OCR for multimodal moderation.

Live moderation117,635,940 scanned
Comment
giveaway_qr.png800 × 800 · PNG · 397 kB
QR codeRemove
Story
match_day.jpg3024 × 4032 · JPEG · 2.7 MB
Identifiable personKeep
Listing
product_shot_final.png2000 × 2000 · PNG · 2.5 MB
BarcodeKeep
Post
unboxing_shot.jpg1440 × 1080 · JPEG · 373 kB
Identifiable personKeep
Direct message
screenshot_0912.png1170 × 2532 · PNG · 1.8 MB
Sensitive personal informationRemove
Story
IMG_0312.jpg1080 × 1920 · JPEG · 498 kB
Bare male chestKeep
Post
team_offsite.jpg4032 × 3024 · JPEG · 2.9 MB
Identifiable personKeep
Direct message
desk_setup.jpg4032 × 3024 · JPEG · 3.1 MB
Tobacco & vapingKeep
Story
happy_hour.jpg2048 × 1365 · JPEG · 671 kB
AlcoholRemove
Listing
poster_final.png1200 × 1600 · PNG · 1.2 MB
WeaponsRemove
Story
clip_thumb_44.jpg1280 × 720 · JPEG · 221 kB
Physical violenceRemove
Profile picture
profile_2024.png512 × 512 · PNG · 163 kB
Identifiable personKeep
How It Works

Know exactly what every image contains.

Send an image by URL or directly as image data. Bodyguard analyzes the visual content and returns every detected classification with its confidence score.

A single image can carry multiple classifications at once, providing a detailed view of its content. With optional OCR, the same analysis can also detect and moderate text embedded within the image.

Analyzed image with the list of detected classifications
Classifications

Detect the visual content that matters.

Bodyguard's image moderation taxonomy covers six categories, from explicit and harmful imagery to regulated content and personal information. OCR, CSAM and QR code detection are add-on options.

Sexual Content

Detect explicit, sexual and suggestive imagery across different levels of severity. Includes signals related to nudity, sexual activity and CSAM.

Violence & Harm

Identify violent, graphic or potentially harmful visual content. Includes signals related to extreme violence, physical violence, weapons and self-harm.

Hate & Harassment

Detect imagery associated with hateful, abusive or harassing behavior. Includes signals related to hate symbols, harassment, bullying and offensive gestures.

Substances & Regulated Activity

Identify visual content involving regulated products and activities. Includes signals related to drugs, tobacco, alcohol and gambling.

People & Identity

Identify people and sensitive information visible within images. Includes signals related to minors, identifiable people and personal information.

Codes & Information

Detect scannable information contained within user-generated images. Includes signals related to QR codes, barcodes and other encoded information.

CSAM Detection

Detect CSAM.

CSAM is the highest-severity classification in Bodyguard's image taxonomy, covering visual content depicting the sexual abuse of minors.

Bodyguard's image moderation technology can detect CSAM as part of the same analysis used to classify other forms of user-generated visual content.

Blurred image flagged as sensitive, with its classifications
OCR detection

Moderate words inside the image.

With optional OCR, Bodyguard extracts text found within an image and analyzes it using the same Text Moderation API used for written content.

One request can therefore combine:

  • Image moderation to classify the visual content.
  • OCR to extract text contained within the image.
  • Text moderation to analyze the extracted words using Bodyguard's text taxonomy.
Diagram: text extracted from an image with OCR, then analyzed with the Bodyguard taxonomy
Performance

Built to moderate at scale.

4B+

Content analyzed

<50ms

Average response time

99.99%

SLA

46

Languages supported

Customization

Image moderation tailored to you.

Go beyond the standard taxonomy with Custom Classifications designed around visual content specific to your business.

Custom Classifications are created on request by the Bodyguard team and tailored to your business and moderation needs.

Adding a custom classification to the moderation taxonomy
More Moderation Products

Moderate more than images.

Text Moderation

Analyze comments, messages and posts with contextual signals and detailed classifications.

Discover Text Moderation

Username Moderation

Analyze usernames to detect inappropriate or undesirable content.

Discover Username Moderation
Trusted by

Protect your image like the biggest brands.

Brands, platforms and public organizations trust Bodyguard to automate their moderation, reduce manual work and improve the quality of online conversations.

VoodooPSGAnkamaM6UbisoftFrance TVReal Madrid CFAudiTF1Petit BateauKonbiniLigue de Football ProfessionnelDecathlonJuventusWizzFC Bayern MunichBrutToulouse FCStream

Ready to moderate visual content at scale?

Bring image moderation and multimodal analysis directly into your platform with Bodyguard.

Book a demo