Developer API for moderating images, video, live streams, text and audio, with 150+ image moderation classes and published monthly plans.
Pick Sightengine if you are a developer who wants moderation with prices on the page, a free tier for testing and live-stream support from the Pro plan. Skip it if you need human review queues or DSA reporting. Its free plan stops at 500 operations a day.
Self-serve plans with published prices that include live-stream and audio moderation, not just images.
Its 150+ image classes, documented class by class in its public docs, give developers fine control that a single safe or unsafe flag does not, while cloud APIs and free endpoints undercut it on simple text and image checks.
Main trade-off
Only classifiers: no review tooling or compliance workflow.
| Plan | Monthly price | What it includes |
|---|---|---|
| Free | Free | 2,000 operations a month, at most 500 a day, Basic content moderation, AI detection, 1 request a second |
| Starter | $29/mo | 10,000 operations a month, then $0.002 each, Visual and text moderation, Videos up to 50MB, no live streams, AI image and video detection, deepfake detection, Email support |
| Pro | $99/mo | 40,000 operations a month, then $0.002 each, Live-stream moderation and videos up to 500MB, Audio moderation, Custom image, video and text lists, Liveness, recapture and age estimation, 10 requests a second, priority support |
| Enterprise | Custom | Custom operation volume, Custom models and classes, Geo-fencing and regional endpoints, SAML SSO, Enterprise SLAs and a dedicated support engineer |
Free: 2,000 operations a month (at most 500 a day). Starter $29 a month: 10,000 operations, visual and text moderation, videos up to 50MB, no live streams. Pro $99 a month: 40,000 operations, live-stream moderation, videos up to 500MB, audio moderation, custom lists, 10 requests a second. Extra operations on paid plans cost $0.002 each. Enterprise is custom, with custom models, geo-fencing and SLAs.
Sightengine is a content moderation API from a French company that launched its first moderation API in 2013. It scores images on more than 150 classes and also moderates video, live streams, text, audio and text inside images, returning a probability per class so developers set their own thresholds.
Image moderation
More than 150 classes, from nudity intensity levels to weapons, gore, hate symbols, drugs and gambling, each returned as a probability.
Video and live-stream moderation
Checks uploaded video frame by frame at a sampling rate you set, and on the Pro plan watches live streams in real time.
Text moderation
Flags profanity, personal information such as phone numbers and emails, and harmful text in chat, comments and profiles.
Audio moderation
Moderates spoken content in audio files, available from the Pro plan.
OCR and QR moderation
Reads text and QR codes inside images and video so spam links and slurs hidden in pictures are caught.
Custom lists and near-duplicate search
Blocklists of known images, videos and words, plus near-duplicate detection, stop re-uploads of content you already removed.
| Tool | Price from | Free plan | Runs on | Rating |
|---|---|---|---|---|
| $29/mo | Yes | Web | 3.9 | |
| Quote | No | Web | 3.6 | |
| Quote | No | Web | 3.8 | |
| Quote | No | Web | 3.9 |
Proprietary
Starter is $29 a month for 10,000 operations and Pro is $99 a month for 40,000, with each extra operation at $0.002. Enterprise is quoted. Watch the unit: running two models on one image uses two operations.
Yes, there is a free plan with 2,000 operations a month, limited to 500 in a day, for testing and proofs of concept. It covers basic moderation and AI detection but not live streams or audio.
Operations are Sightengine's billing unit. Running one model, such as nudity detection, on one image counts as one operation, and each API response says how many operations it used.
Yes, on the Pro and Enterprise plans. Starter handles uploaded video up to 50MB but not live streams.
The service keeps running. Sightengine applies its per-operation rate to the extra usage and bills it a few days later.
Sightengine is a French company that says it launched the first AI-based API for content moderation in 2013 and now serves hundreds of organisations from datacentres in several countries. It is headquartered in France, works remote-first, and its models are its own, built and tuned in-house rather than resold from a cloud provider.
You send an image, a video URL, a live stream, a block of text or an audio file to its API and choose which models to run. Image moderation covers nudity at several intensity levels, violence and gore, weapons, hate symbols, drugs, alcohol, gambling and text or QR codes inside the picture, and returns a probability for each class, so your code decides the threshold. Video is checked frame by frame at a rate you pick, and the same account reaches AI-image, deepfake and AI-music detection, which sit closer to the AI content detectors shelf. When Hokai ran the public moderation demo on 22 November 2026, with no account, its preloaded sample photo scored None at 99% on the nudity scale, with Sea, lake, pool at 60% as context and Bikini at 0%.
It suits developers at marketplaces, dating apps, social apps and photo communities who want to wire moderation in themselves and know the bill in advance. Sightengine publishes plan prices, counts usage in operations (one model on one image is one operation) and charges a flat rate per extra operation rather than stopping the service.
It is a set of classifiers, not a moderation desk. There is no review queue, appeals flow or legal reporting, so a team that needs those in Europe should look at Checkstep, and a platform that wants one large vendor for audio, OCR and child-safety detection can compare Deploy Content Safety. For text alone, a free endpoint from OpenAI may be enough. Hokai keeps these under AI governance tools.
Product facts on this page (pricing, plans, integrations, compliance) are read from Sightengine's own pages and documentation. The verdict is Hokai's reading of them, not the vendor's claim.