Asset 20 8 2
Does AI recommend your business? Run the free check →

Join 15,000 business owners, marketers and entrepreneurs. The Sunday newsletter you'll be annoyed only arrives once a week.

Article

Which Image Moderation API? 10 Options for Apps With Uploads

The best image moderation APIs for most apps are hosted vision APIs that score each upload for nudity, violence and other unwanted content, with full moderation platforms the better fit once you need a human review queue. Decide first what you must block outright and what can wait for review, then pick the type of tool that handles that split.

Pick the wrong image moderation API and you find out in one of two ways: a user posts something vile that sits live on your app for an afternoon, or your support inbox fills with people whose holiday photos were blocked. Both come less from the model than from the decisions around it: which categories you block on sight, which you hold for a person to check, and what you're prepared to let through and clean up later. Make those calls first and the shortlist almost writes itself.

Start with what you must catch

Write down the categories where a single miss is a disaster for your app. For a dating or social app that's usually explicit nudity in profile photos and anything involving minors. For a listings site it might be weapons, drugs or gore in product shots. For a kids' product the list is longer and the tolerance is close to zero. These categories get an automatic block at a low confidence threshold, and you accept that some innocent images will be stopped too.

Then write down what you'd like to catch but could survive missing for an hour: suggestive but not explicit images, rude gestures, offensive text inside an image, low-quality spam. These go to a review queue or get published and flagged for a person to check. The threshold here is higher, because you're trading a few misses for far fewer false alarms that annoy real users.

Last, decide what you'll leave to your community. Keep a report button as the backstop for everything your filters miss, and make sure each report lands somewhere a person will see it the same day.

Once you've got those three lists, check the class names each API returns against them. If a vendor has no class for something on your must-catch list, it's out, however good it looks elsewhere.

Run a side-by-side on your own uploads

Vendor demo images are chosen to look good. Your uploads won't be. So build a test set from real material: pull a few hundred recent images from your own app, including every one your team has removed by hand in the last few months, and label each one yourself as block, review or allow using the three lists above. Mix in some awkward cases on purpose: swimwear, breastfeeding photos, medical images, classical art, a toy gun, a steak on a barbecue. Those are where tools disagree.

Send the same set through two or three shortlisted APIs during their free allowance and put the scores in a spreadsheet next to your labels. For each must-catch category, find the threshold that catches every block image, then count how many allow images it would also have stopped. That number is your false alarm rate, and it's what your users will feel. Note the response time on a normal upload too, because a slow check on a profile photo is a slow sign-up.

If one tool catches everything on your must-catch list with the fewest false alarms, you're close to a decision. If none does, you need a review queue, which pushes you towards a platform or a human service.

Vision APIs you call per image

Sightengine

A moderation API covering images, video, text and audio, with a long list of image classes plus AI-image and deepfake checks.

Hive (now Deploy)

Visual moderation models from the company now trading as Deploy, returning many trust and safety subclasses in a single API response.

ModerateContent

A realtime image moderation API built for websites and apps that accept uploads from their community.

API4AI NSFW API

Cloud NSFW recognition from a small vendor whose sixteen APIs draw on one shared pay-as-you-go balance.

Nyckel NSFW image classifier

Developers who need moderation tuned to their own content can call this pre-trained NSFW classifier and retrain it on their examples.

Amazon Rekognition Content Moderation

For apps already on AWS, a managed API that flags inappropriate images and video using moderation rules you can customise.

Azure Content Safety

Microsoft's service for scoring text and images for harm and flagging items for review, a natural fit inside an Azure stack.

Platforms and services with review built in

Checkstep

A moderation platform scanning images, video and audio, with a review dashboard and policy enforcement layered on top of the automated checks.

Work with me

Want AI doing the heavy lifting in your marketing?

I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.

Moderation API

Rules, review queues and AI scoring across text, images, video and audio, built for product teams who'd rather not code the workflow themselves.

Besedo

Part software, part people: a moderation service combining AI filters with human moderators across text, images and video.

Pure API or full platform

A pure vision API takes an image and returns scores: this much nudity, this much violence, this much weapon. Everything after that, the thresholds, the queue, the appeals, the audit trail, is code you write. That's the right choice for a developer team that wants control and already has an admin panel where flagged items can land.

A moderation platform wraps the scoring in rules, review queues, dashboards and policy enforcement, and often handles text, video and audio as well. You write less code and get a workflow your trust and safety staff can use on day one. Some go further and pair the software with human moderators who review on your behalf, which suits a company without its own moderation team.

The third question is scope. At least one API here lists AI-image and deepfake detection alongside nudity and violence classes, which matters if fake profile photos are a problem for you. If your app generates images rather than only accepting them, our guide to negative prompts for AI images shows how to cut what the generator produces in the first place, though it doesn't replace checking the outputs. Our guide to free AI image detectors covers that check on its own.

Where moderation projects go wrong

The common failure is setting one threshold for everything. A single cut-off either blocks too many harmless photos or lets borderline ones through, and both drive complaints. Set thresholds per category, matched to the three lists you wrote at the start.

The second is moderating on upload and never again. Images get edited, profiles get changed and reports come in later, so plan to re-check when an image is replaced or reported. The third is forgetting the people: whoever reviews the queue will see the worst content your app receives, so blur by default, rotate the work and keep the queue short by tuning thresholds rather than piling everything into review.

The type I'd pick for each situation

If you're a small developer team with clear rules and an admin panel already, start with a pure vision API that publishes its pricing and has a free tier you can test on, and build the thresholds yourself. If you're already deep in AWS or Azure, test your cloud provider's moderation service first; the integration and billing are already in place.

If you've got a trust and safety lead and a growing review backlog, go for a platform with queues and policy rules built in. And if nobody on your team should be looking at harmful images at all, choose a service that adds human moderators to the automated checks, and make sure the contract says who sees what.

Do you make image moderation APIs or content moderation platforms? Tool makers can ask to be considered for a featured place on this list.

Paying per image: what the bill looks like

The pure APIs here that show a price bill on usage rather than seats, while Besedo, the human-plus-software service, doesn't publish a rate at all. The units differ: operations per month on a plan with a charge for each extra operation, images counted in thousands against a prepaid balance, or a fee per call after a monthly allowance. If you add video, ask how frames are counted, since one clip may be billed as many images. If one upload runs through several models, you may pay for each.

A few vendors work the other way. One here asks for a minimum monthly commitment on its commercial plan, and Hive's pricing page, now under the Deploy name, wasn't readable when we checked, so expect to ask for a quote. At the cloud end, Amazon says Rekognition has no minimum fees, licences or upfront commitments, so you pay only for images processed.

For testing, Sightengine's free tier gives 2,000 operations a month, capped at 500 a day, and Nyckel's free plan includes 100 calls a month; API4AI starts you with a small free credit. That's enough to run the side-by-side test. Past the free allowance, the three tools here that quote a per-image or per-call rate sit at roughly $2 to $5 per thousand images on the vendors' own pricing pages this month, before any minimum commitment. To estimate the real bill, multiply your daily uploads by thirty, add any video frames, and ask each shortlisted vendor to price that exact number.

Want your tool on this list? Featured spots and reviews here are paid partnerships. Here's how to partner with me.

Published and maintained by the Lilach Bullock team, covering marketing, AI and business growth.
Your buyers are asking AI who to use. Does it say you?

See for free whether ChatGPT, Claude, Perplexity, Gemini and Google name you, and get the plan to become the answer.

Check my AI visibility →
Sundays only

Get the Sunday newsletter.

One email a week. AI experiments, marketing tactics, and the workflows Lilach is building right now in her own business.

Subscribe free

Let’s get your marketing running on AI.

Book a free 30-minute call

We figure out what you need, where AI fits in, and what working together would look like.

Book the call →

Or take the 30-second calculator

You’ll see the hours and the money quietly leaking out of your week, and the three workflows worth building first.

Take the calculator →

Or grab the free AI resource library

Prompt packs, templates, checklists, and swipe files. The exact tools I build for paying clients. Yours, free.

Get the library →
Keep reading

More from the blog.