Skip to content
Hacker News front page

BFL launches FLUX 3: a single model that generates video, images, and audio together

Flux 3

FLUX 3 is BFL's new multimodal foundation model that jointly trains on images, video, and audio in a unified architecture. Instead of treating each modality separately, it learns from their mutual constraints—sound matching impact, motion obeying mass—to build a representation of the physical world. Video generation is now in early access: text-to-video, image-to-video, and video-to-video, up to 20 seconds with native audio. In early evals, FLUX 3 beats Grok Imagine Video in 69% of comparisons, Runway Gen-4.5 in 77%, and Luma Ray 3.2 in 93%; it leads Kling v3 Pro 60% of the time. Image generation early access opens in the coming weeks. For action prediction, BFL partnered with mimic robotics to test a finetuned version on real production tasks at Audi. The post does not disclose pricing or a general release date.

Why it matters: BFL released FLUX 3, jointly trained on images, video and audio in one architecture, with video generation in early testing, three generation modes, up to 20 seconds and native audio.

Read the original ↗Export Markdown