The product
Twelve Labs builds foundation models that understand video natively: search hours of footage by describing a scene, generate summaries, extract moments. Customers range from media companies sitting on decades of archive to security and sports-analytics operations.
Why now
Text was the first modality to become computable at enterprise scale; video is the largest one that has not. Most organizations cannot even search their own footage. As inference costs fall, the economics of processing video finally work, and a category is forming around it.
What it means for founders
Modality gaps are still open territory. The playbook that worked for text, pick an industry, own its workflow, integrate deeply, is being rerun for video, audio and sensor data with far fewer competitors per niche. If your market has data that is valuable but unsearchable, that is a company.