Why FLUX 3 Trains on Images, Video and Sound — Andreas Blattmann & Dustin Podell, Black Forest Labs
What can a model learn about the world by watching video and listening to it? Black Forest Labs trained FLUX 3 on images, video and audio together. Andreas Blattmann and Dustin Podell join swyx to explain why. A moving camera reveals the far side of an object. A soundtrack gives a second clue to when something happened. A video upscaler has to preserve details across time or the picture flickers. We trace the decisions behind FLUX, from the early latent diffusion work to the training data behind Kontext and the tradeoff between diffusion and autoregression. At the end, Dustin shares a use he didn’t see coming: people using FLUX to restore old photos and showing the results to their families. This episode is sponsored by Deepgram. Try Deepgram for voice agents Show notes BFL explains what a still image, a moving camera and a soundtrack each contribute during training. FLUX 3 launch and research overview Watch and listen to the results of generating video and sound together. FLUX 3 Video demos See how the lab teaches a model to make a requested change while preserving the rest of an image. FLUX Kontext examples Andreas Blattmann co-created Stable Diffusion and co-invented the latent diffusion method behind it. Guest bio · Original research Dustin Podell led FLUX Kontext and was the lead author of SDXL. Dustin’s project credits · FLUX Kontext · SDXL The first Latent Space essay grew out of the Stable Diffusion release. Multiverse, not Metaverse Chapters 0:00 Host introduction3:39 Cold open4:23 Black Forest roots9:23 Dustin’s path17:04 Visual intelligence27:38 Inside SDXL33:59 Learning motion39:57 Flow matching46:57 Starting BFL50:36 Deepgram sponsor break51:43 Early use cases for FLUX54:58 Editing with Kontext1:03:15 Training data1:07:32 Testing an edit1:13:12 Autoregression1:20:50 World action models1:30:08 Robotics in 20271:32:16 Who uses FLUX?