Linum

Linum is a two person research lab that trains generative video models from scratch. Its stated aim is animation for anyone with a story, with the controls for artistic intent built into the model training itself rather than added through fine tuning or complicated workflows. The company says it is backed by Y Combinator. Its current release is Linum v2, a text to video model with 2 billion parameters. The weights are available to download under the Apache 2.0 license, so users can run the model themselves. The home page shows example videos generated from detailed text prompts, such as an animated goat playing guitar or honey poured over pancakes. A field notes section publishes research, including a series called Road to Linum v3 on text to image training and data filtering for video pre training. The audience is researchers, filmmakers and animators who want an open video model. The site shows no hosted app, subscription or pricing, and the model weights are the published offering.