Megvii Face++ Review

Face++ is run by Megvii. Its Facial Recognition API group covers Face Detect (which returns face tokens), Face Compare (1:1), Face Search (1:N against a FaceSet you build), Face Analyze, dense landmarks, attributes, emotion recognition, beauty score, gaze, skin analysis and 3D face reconstruction. The wider platform adds identity products (KYC, liveness, deepfake detection), body and gesture recognition, OCR and image beautification. Offline SDKs such as the Face Compare SDK run on device.

You sign up, get one free API key, and call REST endpoints. The free key shares QPS capacity with other users; paid use is either pay as you go from a prepaid balance or a daily or monthly plan that buys guaranteed queries per second. Buyers in the US should know that Megvii Technology was added to the US Commerce Department's Entity List on October 9, 2019, per the Federal Register.

Face++ lists four models: Free, Pay As You Go (top up a balance, charged per call, including failed calls), Daily/Monthly Plan (prepaid QPS capacity, unlimited calls in the period), and licensing for SDKs.

What that means: 1,000 face comparisons cost 1,000 x $0.002 = $2.00. A month of 1 guaranteed QPS ($1,000) allows up to 1 x 86,400 x 30 = 2,592,000 calls, which beats pay-as-you-go only past 500,000 calls ($1,000 / $0.002).

Free: one API key per account, no limit on total usage, 3 QPS shared across the Facial Recognition group (1 for Dense Facial Landmarks). SDKs: Face Landmark SDK online licences cost $4,000 a year for 1,000 licences, $10,000 for 10,000 and $20,000 for 100,000, per platform; offline SDKs (including Face Compare SDK) are priced by sales. Balances do not expire.

Prices checked on the vendor's pricing page on November 12, 2026. See Face++ pricing details.

What Face++ lists under Facial Recognition and Identity:

Face++ suits developers who want a face API with self-serve sign-up, a free key for testing, per-call prices around $0.002, and attribute APIs (emotion, beauty, skin) that Microsoft has retired or restricted. The QPS plans work for steady high volume where a per-call bill would be unpredictable.

Teams subject to US export controls, or that need a Western-hosted service for procurement, should look at Amazon Rekognition or Azure AI Face. To keep images on your own hardware, self-host CompreFace or DeepFace.

Independent reviews are scarce. FAUN.dev scores Face++ highest on pricing transparency and maturity and lowest on custom models, notes that face tokens and FaceSets cannot be exported, and flags that the hosted API runs on Megvii's servers while the company is on the US Entity List. Banuba's comparison, written by a competing vendor, credits Face++ for its landmark and attribute research.