# SightRadar vs Google Cloud Vision API

> Google Cloud Vision detects faces — it does not recognise them. Google's own documentation states that "specific individual Facial Recognition is not supported": you get bounding boxes and facial attributes such as emotion likelihood or headwear, with no collections, no enrolled identities, and no similarity search. So if your product needs to answer "which photos is this person in?" or "is this the same person?", Cloud Vision cannot do it at any price. SightRadar is a dedicated face-recognition API — index, search, compare, detect — with calibrated similarity scores, per-tenant isolation, and flat per-photo pricing ($0.00093/photo).

Source: https://sightradar.com/compare/google-cloud-vision

## Bottom line

This is a capability comparison, not a price one. If your product needs to know WHO is in a photo — find every photo a guest appears in, verify a selfie against an ID, catch a duplicate account — Cloud Vision cannot do it at any price, because Google does not offer individual facial recognition. SightRadar is built for exactly that job. If you only need to know that A face is present, or you want labels and OCR too, use Cloud Vision — the two are complementary rather than competing.

## Headline verdicts

### Face recognition (identity) — winner: SightRadar
- SightRadar: Index faces into collections, search with a selfie, and compare 1:1 with calibrated similarity scores.
- Google Cloud Vision: Not supported. Google's docs state specific individual facial recognition is not available — detection and attributes only.

### Price — winner: Tie
- SightRadar: $0.00093/photo real-time, $0.00062/photo batch — for face RECOGNITION: index, search, compare.
- Google Cloud Vision: $0.0015/image for Facial Detection (1,001-5,000,000 units/month), billed per feature applied. Cheaper per call than recognition — but it is a different, lesser operation, so the rates are not comparable.

### Web UI — winner: SightRadar
- SightRadar: A console to browse collections, run searches in a playground, manage keys, and see per-operation usage and cost.
- Google Cloud Vision: The Google Cloud console is a general-purpose cloud UI — there are no face collections to browse, because collections do not exist.

## Where SightRadar wins

- Cloud Vision cannot identify people. Google's documentation states plainly that specific individual facial recognition is not supported, so selfie search, find-my-photos, and identity verification are simply not buildable on it.
- No collections and no faceprints means no 1:N search: there is nothing to enrol a person into and nothing to search against.
- SightRadar returns calibrated similarity scores you threshold yourself, which is what makes verification and dedup decisions defensible.
- A Rekognition-compatible API plus first-party Python and Node SDKs, so an existing face-recognition integration ports over instead of being rewritten.
- A real console: browse collections, run a search in the playground, and see the exact credit cost of every operation.

## Where Google Cloud Vision is the better fit

- If you never need identity and want broad image understanding — labels, OCR, landmarks, logos, object localisation — from one API, Cloud Vision is genuinely the better and cheaper tool, and its face-attribute detection (emotion likelihood, headwear) is strong. Plenty of teams correctly use Cloud Vision for general image analysis and a dedicated recognition API for identity.

## Feature-by-feature

### Core capability

| Feature | SightRadar | Google Cloud Vision |
| --- | --- | --- |
| Identify who a face belongs to (1:N search) | ✅ Search a collection with one selfie, ranked by similarity | ❌ Not supported — detection only, no enrolled identities |
| 1:1 face verification (same person?) | ✅ Compare returns a calibrated score you threshold | ❌ No comparison operation for individuals |
| Face detection + quality gating | ✅ Detect locates and quality-gates faces without storing anything | ✅ Detection with attributes (emotion likelihood, headwear) is its strength |
| Persistent face collections | ✅ Collections are the core primitive; delete any time | ❌ No concept of a collection or a stored faceprint |
| Broad image understanding (labels, OCR, logos) | ❌ Out of scope — SightRadar does face recognition only | ✅ Labels, text/OCR, landmarks, logos, object localisation |

### Getting started

| Feature | SightRadar | Google Cloud Vision |
| --- | --- | --- |
| Rekognition-compatible API | ✅ Drop-in IndexFaces / SearchFacesByImage / CompareFaces shapes | ❌ A different API surface entirely |

### Web UI & visibility

| Feature | SightRadar | Google Cloud Vision |
| --- | --- | --- |
| Per-operation cost visibility | ✅ Exact credit cost per operation in the console | ➖ Cloud Billing reporting across all GCP services |

### Pricing

| Feature | SightRadar | Google Cloud Vision |
| --- | --- | --- |
| Billing unit | ✅ Per photo processed — $0.00093 real-time, $0.00062 batch | ➖ Per feature per image; each feature applied is its own billable unit |

Competitor pricing changes over time — verify the current rate at https://cloud.google.com/vision/pricing. SightRadar's rates are at https://sightradar.com/pricing.

## Try SightRadar

Self-serve sign-up, pay only for what you use, AWS Rekognition–compatible API, ready-made Python & Node SDKs, and a real web console. Start free at https://sightradar.com/login — docs at https://sightradar.com/docs.