Ask any video a question.
Get back the exact second.
WAVE Vision gives agents eyes — and bills like every WAVE product: pay per call, not a license. It chunks a stream or recording, captions every moment with a vision model, embeds it, reranks it, and composes grounded, timestamped answers. No NIMs to self-host. No GPU cluster to babysit. An unpaid call gets a 402 challenge; pay and retry — no subscription required.
Pricing — metered per call
Index · Summarize
$0.05 / analyzed-minute — billed on the media you actually analyze, tier-down to $0.03.
Search · Ask
per query — a grounded, cited answer costs a query, not a subscription.
The four calls
1. Index
POST /v1/vision/index — embed your caption chunks, get back a reusable index.
2. Search
POST /v1/vision/search — ask in plain language, get back the timestamped moments.
3. Ask
POST /v1/vision/ask — a grounded answer, cited to the moment it's built from.
4. Summarize
POST /v1/vision/summarize — a summary focused on your question.
Agents have eyes
An agent discovers this surface at /llms.txt and calls it unattended — same Bearer key, same x402 wallet, same meter as every other product on the WAVE Protocol Plane. Video intelligence stops being a project you stand up and becomes one composable call.
preview · cloud-hosted + metered on the WAVE protocol plane · per-call