Best fit
- large sets of frames
- scene tagging
- factual captioning
- video metadata enrichment
Recommended stack
- model: ClipTagger
- small related bundles: background jobs or group jobs
- large queues: batch
Workflow
- decide the frame sampling strategy
- choose whether the job is small-batch or large offline
- run frames through ClipTagger
- aggregate frame outputs into your higher-level video result