A media archive needs structured topics and events from videos, not just a raw transcript. Which capability should be evaluated?

1. Text-to-speech only
2. A file checksum alone
3. Multimodal information extraction
4. DNS resolution