Are you an LLM? Read llms.txt for a summary of the docs, or llms-full.txt for the full context.
Skip to content

Face Swap Detection

The earlier face-swap detector for videos.

SlugIDStatusTaggedModel version
faceswap-detection4DeprecatedNov2

What it detects

DetectsDoes not detect
Face-swapped deepfakes in video, checking each person who appearsVoice cloning, or edits that don't involve faces

Request

SlotAccepted contentType
originalvideo/mp4, video/mov, video/webm
{
  "taskTypeId": "4",
  "inputs": [
    { "slotName": "original", "fileName": "interview.mp4", "contentType": "video/mp4", "sizeBytes": 5242880 }
  ]
}

Limits

  • File size: up to the account upload limit, 10 MB by default.

Result

This task type counts faces instead of returning a score, so its results have no fakeConfidencePercent or tag.

FieldTypeDescription
isFakebooleantrue if any detected person's face is swapped
deepFakesintegerHow many of the detected people have a swapped face
totalFacesintegerHow many distinct people were detected
modelVersionstringVersion of the model that produced the result
{ "isFake": false, "deepFakes": 0, "totalFaces": 2, "modelVersion": "v2" }

If no faces are found, isFake is false and both counts are 0.

Output artifacts

kindFileDescription
resultresult.jsonPer-person verdicts and confidence, with face bounding boxes by frame

Migrating to ekyc-deepfake

  1. Change taskTypeId from 4 to 11. The input slot and accepted formats are the same.
  2. Read fakeConfidencePercent and tag instead of deepFakes and totalFaces. See Reading results.
  3. Re-check any thresholds you apply: scores from different models aren't comparable.

Cost and availability

  • Fixed price per job, set by your plan. The exact amount is returned as cost when you create the job.
  • Available on the cloud API until its sunset date.