Sieve (YC X25) Is Hiring Engineers to build video datasets for frontier AI
A dataset suite pushing the frontier of video generation, human avatars, and world models.
Video Generation
GeneralCinematic
Human Avatars
Talking Head HumansFully Body HumansAnimated Characters
World Models
Real World EgocentricRendered Egocentric
500K hours of high quality, diverse video clips.
We offer additional packaged datasets not listed here.
Contact us to request a sample or explore more options.
High Quality
Purpose-built video understanding models paired with human QA help find just the highest quality, training-ready data.
Unparalleled Scale
Our growing library consists of thousands of petabytes of video data.
Extreme Diversity
Video is collected from a variety of public, private, and synthetic sources.
Next-Gen Complexity
New data shapes to unlock new model capabilities (paired, time-synced, conversational, and more).
