Unique Frame ID: less noise, better search
Most video does not change much from one second to the next. Analysing every frame creates piles of repeated tags that bury the results you want.
Unique Frame ID finds the frames that matter and analyses only those.
The result
- 87% less metadata noise, so search results are more relevant. In a 12 minute test video, 97 of 719 frames were kept, 87% fewer than tagging one frame per second.
- Faster processing, because there is less to analyse.
- Lower cost, because you analyse fewer frames.
In simple footage, such as a single-camera interview, the reduction can be higher still.
How it works
12 minute video, one frame per secondExtracting frames
719frames to analyse
0%fewer frames
Real test result: 97 of 719 frames kept. Each square stands for about 10 frames. Colours show scenes.
- Frames are extracted from your video, typically one per second.
- Every extracted frame is scored for its importance.
- It is scene aware and semantic, so it understands what is in each frame and where scenes change.
- The best and most relevant frames are kept, and repeats are dropped.
- The models analyse only those frames, and the results map back across the full timeline.
FAQ
Does it miss things?
It is designed not to. We score every frame we extract for its importance, then select the best and most relevant ones. Because it is scene aware and semantic, it captures the important moments and drops only the repeats.