Queryable, verified video

Stop searching your video. Start acting on it.

A search that just returns timestamps leaves you with a list. AT-1 turns the query into an action: find the moment, pull that exact frame or clip out as a standalone verified artifact without rehydrating the rest, and prove it was never altered.

We carry the standard codec’s frames (H.264 / AV1 / DICOM) — we never re-encode the pixels. The win is the queryable + verified + addressable layer they omit.

Find it

Query by time, motion/activity, scene-cut, or search-by-image — all over an embedded per-frame index, with no full decode. Jump straight to the moment in a ten-hour file.

Pull it out

Extract that exact frame or clip into a new, standalone artifact — built only from the frames you matched. The other ten hours are never read or rehydrated.

Prove it

Every frame is SHA-256 hashed. Verify byte-exactness, and a single altered byte is detected and pinpointed to the exact frame. Redact PII and re-stamp the chain of custody.

# find the activity, pull it out as a new verified clip, prove it
at1 media query camera.at1vid motion 1.2
at1 media clip  camera.at1vid event.at1vid 21,22,23,24,25
at1 media verify event.at1vid  # byte-exact; tamper -> located
at1 media redact camera.at1vid clean.at1vid 0,1 # PII, re-stamped

Where the layer is the pain

Medical / PACS

Lossless is mandated, archives are exabytes, and the need is addressable + integrity-verified retrieval — not smaller pixels.

Surveillance / defense

“Jump to the frame with the plate” + “prove this footage wasn’t tampered with.” Provenance and query, not ratio.

Media asset management

Search a video lake by content, address any shot, verify masters — one signed, queryable artifact.

Same primitive, every data type: the result of a query is always a new verified, re-queryable artifact — a video clip, a columnar --extract, a genomic --region sub-cohort, a table or bundle subset — read only from what matched, never rehydrating the source.