Check an image before submitting it to Create Face. Returns an overall pass/fail result, structured checks, and diagnostic details without creating a face.

Send the image bytes as base64 in imageBase64, without a data URL prefix, and specify the image's mimeType. This endpoint takes image data, not a train_image_url.

Set model to the same Phoenix model you will use as model_name when creating the face. Validation defaults to phoenix-4; send "model": "phoenix-4.5" to use Phoenix-4.5 checks. See Phoenix-4 Image Requirements and Phoenix-4.5 Image Requirements.

HTTP 200 means validation completed, not that the image passed. Read passed for the overall result. Each entry in checks has an id, a display label, and its own passed value. Entries with passed: false identify what the user should fix. The checks and details vary by model.

The JSON request body must not exceed 30 MiB. The base64 string is limited to 28 MiB of characters.

Legacy: POST /v2/replicas/image-validation remains supported. Its diagnostic details use pose_suitable_for_replica where this endpoint uses pose_suitable_for_face; check IDs are unchanged.

{
  "imageBase64": "<base64-encoded-image>",
  "mimeType": "image/jpeg",
  "model": "phoenix-4.5"
}