Picsart just switched on FLUX 3, the new multimodal frontier model from Black Forest Labs , and it is one of the biggest model drops the platform has shipped this year. Announced by the Picsart team on August 6, the integration brings FLUX 3's headline trick to everyday creators: up to 20 seconds of generated video with sound created alongside the visuals, starting from a text prompt, a single image, a reference clip, or a set of defined keyframes. Picsart was one of the handful of platforms named in Black Forest Labs' original FLUX 3 reveal in late July, alongside Canva, Krea, and Freepik, and it is now among the first to put the model in front of a mainstream audience. 20 seconds of video with the audio baked in Most AI video tools generate silent clips and leave you to hunt for music and sound effects afterward. FLUX 3 was trained on images, video, and audio inside a single unified architecture, so every clip comes out with native sound: spoken dialogue, sound effects tied to what is happening on screen, and ambient noise that matches the scene. Black Forest Labs says the model is particularly strong at associating sounds with physical events, so a door slam or a splash actually lands where it should. Clips run up to 20 seconds in a single generation, in HD or Full HD, which is enough for a complete short-form beat rather than a 5-second teaser you have to stitch together. Four ways to start: text, image, clip, or keyframes The Picsart integration covers the full spread of FLUX 3 Video's input modes. You can generate from a plain text prompt, animate a still image, feed in a reference clip for video-to-video restyling or continuation, or define keyframes to control exactly how a scene opens, transitions, and ends. That last option is the sleeper feature for editors: keyframe control gives you directorial say over transitions without manual compositing, which has traditionally been the hardest part of making AI clips feel intentional. Carry a character from one sc