Extract a face embedding from an image.
Public URL or base64 data URI of the image.
Face embedding with 512-dim vector and optional bounding box.
Compare two face embedding vectors using cosine similarity.
First embedding vector.
Second embedding vector.
Minimum similarity to consider a match (default 0.6).
Comparison result with similarity score and match flag.
Extracts face embeddings via the Replicate API using InsightFace.
Sends a prediction request with the provided image URL, polls until the prediction completes, then parses the embedding vector from the response.