In the /v1/avatar_details API we return a key avatar_processing_results in the response payload. This returns a list of ProcessingResult objects each of which has two attributes:

  • check_name: A string representing the name of the check. It can be one of the following:
  • check_status: The status of the check which can have one of the following values

​
check_name

Verification check nameDescriptionHow to Fix
no_faces_checkNo face detectedPlease ensure there is a face in frame at all times.
multiple_faces_checkMultiple faces detectedPlease ensure there is only one face in frame at any time.
blur_checkBlurry video detectedPlease upload a clearer video with better resolution and focus.
bbox_iou_area_checkSudden & fast facial movements detectedPlease record again without any sudden movements, video cuts or scene changes.
face_within_constraints_checkFace outside the constraintsPlease make sure your face is in the recording frame at all times.
min_bbox_size_checkFace detection too smallPlease move closer to the camera for better face detection.
occlusion_checkBlocked face detectedPlease record again without covering your face or mouth.
intra_vid_emb_checkMultiple people detected in videoPlease make sure only 1 person is present in recording.
db_emb_checkPublic figure detected in videoPlease record again without any public figures.
landmarks_extractedFace and lips not detected clearly throughout the videoPlease record again by ensuring that face and lips are visible in all frames of the video.
landmarks_score_checkFace moves too much, lips not detectedPlease record again by ensuring that face and lips are visible in all frames of the video.
syncnet_checkAudio and lipsync not matching in the videoAudio and lipsync are not matching in the video. Ensure synchronization between audio and lip movements.
speaker_diarizationMultiple speakers detected in the videoPlease record again with a single speaker in the video.

​
check_status

check_statusDescription
SUCCESSCheck passed successfully
PENDINGNon mandatory check may still be in progress
FAILURECheck failed
ERRORSystem error occurred during check
DEPENDENCY_FAILEDDependent check failed
INVAILD_VIDEO_TYPEUnsupported video format

In the /v1/avatar_details API we also return a key avatar_verification_failure_metadata in the response payload. This returns a list of AvatarVerificationFailureMetadata objects each of which has three attributes:

  • fps: The fps of the processed video:
  • bboxes: A dictionary of time in milliseconds mapped to a list of bounding boxes (it can have multiple bounding boxes in case of multiple faces detected in frames)
  • failures: A dictionary of check_name mapped to its corresponding AvatarVerificationFailureDetails objects. Each of this object has the:

    • reason: The failure reason as per the descriptions given table above

    • fix: The how to fix message for that error as per the given table above

    • info: In case of the following frame type checks, we will receive the list of frame ranges in the info key that violates a given check.

      ## Frame type checks

"no_faces_check"
"multiple_faces_check"
"blur_check"
"face_within_constraints_check"
"min_bbox_size_check"
"occlusion_check"
"db_emb_check"
"landmarks_score_check"

      Sample info object:

      {
  "fps": 25,
  "bboxes": {
    "0.0": [
      [
        989,    # x1
        449,    # y1
        1258,   # x2
        786,    # y2
        1       # confidence
      ],        # face 1
      [
        1363,
        557,
        1494,
        732,
        0.9999390840530396
      ]         # face 2
    ],
    "40.0": [
      [
        987,
        451,
        1256,
        786,
        1
      ]
    ]
  },
  "failures":{
	  "landmarks_score_check": {
	    "fix": "Please record again by ensuring that face and lips are visible in all the frames of the video",
	    "info": [
	      [
	        21840,
	        22600
	      ],
	      22680
	    ],
	    "reason": "Face moves too much, lips not detected"
	  }
	}
}
Avatar Video GenerationSound Effects