A media file you can query, address, and verify.
Most media tooling makes you decode the whole file to do anything with it. AT-1 Media wraps a video into a queryable, verified container: address any frame, find a scene, extract a clip, and prove no frame was tampered with, touching only the bytes a request needs.
Address any frame
Jump to frame N (or a timestamp) and pull just that frame: no decode-the-whole-file. The container is indexed, so reads touch only what you ask for.
Query & extract in place
Find scene cuts, query by content, and extract a clip without unpacking the entire video: the same query-don't-decompress idea AT-1 applies to data.
Verify & pinpoint tamper
Every frame is integrity-checked; if a frame was altered, verification names exactly which one. Provenance you can hand to a third party.
Works on real camera files
Validated on real camera MP4: find → extract → verify. It's a layer on top of standard media, not a new codec you must adopt everywhere.
Finding the frame is the easy half
Plenty of tools will tell you which frame matched. The question that decides whether it is useful is what happens next. If answering it means rehydrating the whole video, you have moved the work rather than removed it, and the thing you hand over is a copy of everything with a timestamp attached.
So the operation that matters is not the search. It is pulling the matched frames out as a new, independently verifiable artifact while the rest of the source is never read.
One run, end to end
Real camera footage, 100 frames at 1280x720, not a synthetic fixture.
| Source | jellyfish_cam.mp4, 64 MB |
| Query | motion activity, no decode. Bursts at frames 21-25, 47-50, 71-76 |
| Extract | 15 frames into a new 10 MB verified container |
| Source read | only those frames. The other 85 were never touched |
| Verify | independent check passes on the new artifact alone |
| Tamper | detected and located to the exact frame |
The clip is built by copying payload bytes, so nothing is re-encoded and the pixels that come out are the pixels that went in.
What you can ask it
- By time, by scene cut, by motion activity, by similarity to another frame, or by an image you supply. All of it reads the index, not the video.
- Pull one frame out as a real PNG, reading only that frame's bytes.
- Blank frames that contain personal information and re-stamp the chain of custody, so a redacted copy is still provably a copy of the original.
What it is not
It is not a codec, and it will not make your video smaller than H.264 already does. It wraps frames without re-encoding them, which is what keeps extraction byte-exact and also means the compression is whatever your encoder already achieved. If raw ratio on natural video is the problem you are solving, this is the wrong tool and we would rather say so here than after a pilot.
Where we're honest about the win
We don't beat H.264 / DICOM on raw compression ratio, and we don't pretend to. The win is the layer: a verified, addressable, queryable container around your media so you can find, extract, and prove, without decoding everything. For natural video the codec stays standard; AT-1 adds the index, the integrity, and the addressability.