Black Forest Labs video model, which I posted about when it was first teased on their site back in August 2024, is about to arrive in the form of a multimodal Flux 3.
A placeholder page was briefly up yesterday at http://bfl.ai/models/flux-3 before being quickly taken down. It read:
'A breakthrough in control, realism, and world understanding — one multimodal model generating image, video, audio and action.'
So, it appears the Flux 3 model is fully multimodal. I can tell from my timeline that some people have been granted early access, and are posting gens, but they appear to have been told to be mysterious avout the model being used, as none of them name Flux 3. One exciting thing from those posts, and the one below, is that Flux 3 appears capable of 20 second video generations, which would put it ahead of nearly everyone in the industry. Hopefully more news soon.
There were rumors a long time ago that this was the video model Elon was going to use for Grok, before he decided to make his own with Imagine. Flux was also used by Grok to generate images before Imagine image was ready.