Verification (1:1)

1:1 Verification answers the question: "Are these two faces the same person?" It compares two face images (or one image against an enrolled face ID) and returns a similarity score between 0 and 1.

Use verification when you need to confirm that the person presenting themselves matches a known identity, for example re-authentication, document-to-selfie matching, or access control.

Compare two images

The simplest form: submit two face images and get back a similarity score. No collection or enrollment required.

curl -X POST https://api.livexface.com/api/v1/compare \
  -H "X-API-Key: YOUR_API_KEY" \
  -F "[email protected]" \
  -F "[email protected]"

{
  "success": true,
  "data": {
    "match": true,
    "confidence": 0.91,
    "thresholdUsed": 0.45
  }
}

Verify against an enrolled face

Compare a probe image against a specific enrolled face ID in a collection. More efficient than comparing two raw images since the enrolled face reference is already stored.

curl -X POST https://api.livexface.com/api/v1/collections/COLLECTION_ID/verify \
  -H "X-API-Key: YOUR_API_KEY" \
  -F "[email protected]" \
  -F "face_id=dc3e7a57-22d5-4271-b707-213e49c8fe54"

{
  "success": true,
  "data": {
    "match": true,
    "confidence": 0.93,
    "thresholdUsed": 0.45,
    "faceId": "dc3e7a57-22d5-4271-b707-213e49c8fe54"
  }
}

Verify against a reference image

Instead of naming an enrolled face, you can send a second image and compare the two directly. Pass reference_image in place of face_id — the request needs one or the other.

POST /api/v1/collections/{collection_id}/verify
X-API-Key: lxf_a1b2c3d4_e5f6g7h8i9j0k1l2m3n4o5p6
Content-Type: multipart/form-data

image=@probe.jpg
reference_image=@enrolled.jpg

Response fields

Choosing a threshold

The default threshold of 0.40 is calibrated against LFW: it accepts 96.99% of genuine pairs at a false-accept rate below 1e-5. You can override it per-request with the threshold parameter:

• Balanced (0.40, default): Good for most identity verification use-cases.
• High security (0.45–0.50): Fewer false accepts, at a real cost in false rejections — 0.50 drops acceptance of genuine pairs to 93.76%. Use for access control or financial verification.
• Permissive (0.30–0.35): Catches more borderline matches. Suitable for photo deduplication or fuzzy matching.

Raising the threshold well above 0.50 buys almost nothing: the false-accept rate has already bottomed out by then, so the extra strictness only rejects genuine users. At 0.70 the API accepts just 41% of genuine pairs.

# Override threshold per request
image=@probe.jpg
face_id=dc3e7a57-22d5-4271-b707-213e49c8fe54
threshold=0.50

Try it live

Enter your API key, the collection ID, and the face ID you got back when you enrolled the face, then upload a probe image.