Flux 3 produces video with native audio — a first for the industry!

Black Forest Labs just launched Flux 3, their first multimodal model. They trained it on a combination of images, videos, and audio. It generates up to 20 seconds of 720p video with integrated audio. That means no need for an audio overlay process. You can use it for text to video, image to video, and even video to video. Plus, the dialogue in the resulting video can be in different languages.

According to their tests, Flux 3 beat out Luma Ray 3.2 in 93% of their head to head comparisons. While the results aren’t quite as impressive compared to Seedance 2.0 or Gemini Omni Flash (about 52%), there are no independent tests yet. Flux 3 Video is now available, while Flux 3 Image will be available soon in early access form. There will also be an open weight version of Flux 3 Dev. This is a major milestone for AI video because it means we can finally get native audio in our video. Before this, we had to use separate models to add audio. If Flux 3 lives up to the hype, then we’ll be entering a new age of AI video where you can just input one prompt and get all the information you need.