Skip to main content
POST
Search across keyframe thumbnails in a batch by describing what you’re looking for. Uses CLIP text-to-image similarity to match your description against extracted video frames. Cost: 0 credits per query
Batches created with POST /v1/batch are not keyframe-indexed yet, so this endpoint returns total_results: 0 for them rather than an error. See the Visual Search Guide before building against it.
Visual search is in beta. It works best for broad visual descriptions. For details and example queries, see the Visual Search Guide.

Path parameters

string
required
UUID of the batch to search. Must be completed and have keyframe embeddings — a batch with none returns an empty result set.

Request body

string
required
Text description of visual content to find (e.g. "outdoor establishing shot", "person at desk").
integer
default:10
Maximum number of results to return (1-50).

Result fields

Error responses