const res = await client.moderation.check({
entity_type: "listing_photo",
entity_id: "photo-8231",
entity_creator_id: "user-412",
moderation_payload: {
images: ["https://cdn.app.com/8231.jpg"],
},
config_key: "marketplace",
});
// res.recommended_action -> "flag"Image moderation API, scored in 300ms.
Every listing photo, profile picture, and upload scored against your policy without slowing the sender down.
New uploadsLive
Sent to review2 images -> media queue
Moderating user-generated content at scale.
- 2B+images moderated/month
- 40msAPI latency for image
- 100%recall on extreme violence, gore and hate symbols
Explicit, violent, and unsafe images, with thresholds you set.
Set a confidence threshold per category, so you decide what gets removed and what goes to review.
#CategoryWhat it catchesThreshold
01CSAM supportDetect child sexual abuse material in uploaded images and act on it under your policy.Always detected
02Explicit content and non-explicit nudityExplicit nudityExplicit sexual activitySex toysIntimate body partsIntimate kissing90
03Swimwear and underwearRevealing swimwearUnderwear95
04Violence and visually disturbing contentWeaponsPhysical violenceGraphic violenceSelf-harmGore and bloodCrashes75
05Drugs, tobacco, and alcoholDrug productsParaphernaliaTobaccoAlcoholic beveragesAlcohol use85
06Rude gesturesMiddle fingerOffensive gestures90
07GamblingCards and diceSlot machines85
08Hate symbolsExtremist symbolsHate group imagery70
Catch the scam hidden in the screenshot.
OCR reads text inside images in 50+ languages, such as phone numbers, payment handles, and scam offers, and runs it through your text policy. OCR hits are labeled separately in the queue.
OCRScamPlatform bypass
Stop the same image from coming back.
Duplicate detection matches identical images by content, not URL, so re-uploads and spam waves get caught.
A review queue built for trust & safety teams.
Flagged images land in a dedicated media queue with user history, bulk actions, and an audit log.
ModerationText queue128Media queue42Users9RulesAudit log
Media queue42 pending3 selectedRemove all
Moderate any image, from any platform.
Send any image to the Check or Analyze API, by URL or raw bytes: listing photos, profile pictures, and user uploads.
Stream moderation FAQs.
An image moderation API checks user-uploaded images against your policy and returns a decision, such as allow, flag, or remove. Stream scores each image by category with a confidence score, and reads any text inside it with OCR.
Stream detects explicit content, nudity, swimwear and underwear, violence, disturbing content, drugs, tobacco, alcohol, rude gestures, gambling, and hate symbols, with a confidence score for each.
Yes. OCR extracts text from images and checks it against your text policy, so a phone number or payment handle in a screenshot gets flagged.
Yes. Each category has its own confidence threshold, so you can remove high-confidence hits and send borderline ones to review.
Yes. Send images to the Check or Analyze API by URL or as raw bytes.
More content types, same policy.
Send us a sample of your flagged images.
See how Stream scores them against your current tool, at no cost.