I've been building a personal project and wanted to check with the community on multi-modal inputs since I can't find a lot of material around this online. Ultimately I'm trying to build something that can ingest massive length (almost like a full stream - 6-10 hours) and accurately do multimodal an
