WAN 2.7: the most controllable AI video model, now on Picsart

WAN 2.7 is now available to 130M+ creators on Picsart — one of the largest creative platforms in the world, where you generate, edit, and publish in one place. Feed up to 5 reference images, lock your first and last frame, edit generated videos with plain language instructions, and take every output straight into Picsart's editor to refine, layer, and publish. No downloading, no switching tools. Available in AI Video Generator, Fullscreen AI Video Generator, Flow, and Playground. Storyline and Aura integration coming soon.

WAN 2.7 is the next iteration of the WAN model family, building on WAN 2.6 already integrated in Picsart Flow. The upgrade is visible immediately: sharper textures, more stable lighting, cleaner detail in skin, fabric, and hair. Motion dynamics feel physically timed — body movement has natural weight, and camera behavior feels intentional rather than algorithmic. Character consistency holds across multiple shots: same face, same clothing, same props. First-pass quality outputs at 1080p. It supports text-to-video, image-to-video, first-to-last frame generation, and multi-reference video with native audio across 5, 10, and 15-second durations in all popular aspect ratios.

First and last frame control lets you lock your opening and closing shot — WAN 2.7 generates everything in between, giving you control over the narrative arc of every clip. The 9-grid image-to-video system accepts 9 reference images in a 3×3 grid (multi-angle shots, sequential poses, scene variants) and uses all of them to keep the subject stable, the pose accurate, and the composition tight — this is how you get consistency that previous models couldn't hold. Instruction-based video editing lets you fix a video that's 90% right: tell it "change the background to a rooftop" or "make it sunset lighting" in plain language, and it edits without regenerating from scratch. Video recreation takes an existing video and generates variations while preserving the original motion and pacing — same choreography, different style or setting. Subject and voice reference locks both face and voice in one workflow, so the character looks and sounds consistent from a single input.

WAN 2.7 is the default model in Picsart's AI Video Generator and Fullscreen AI Video Generator, and is available in Flow and Playground. The generation is the starting point, not the end — take any WAN 2.7 output straight into Picsart's editor to refine, layer, adjust, and publish. No downloading, no switching tools. WAN 2.7 lives inside one of the largest creative platforms in the world, where the full workflow from generation to publishing happens in one place. Integration with Storyline and Aura is coming soon as a fast follow.

WAN 2.7 gives creators more control than any other video model available today. Five simultaneous reference inputs (up from 2 in WAN 2.6), 9-grid multi-angle consistency, first-to-last frame narrative control, instruction-based editing without regeneration, video recreation that preserves motion, and subject plus voice locking in one workflow. The visual quality speaks for itself — sharper textures, more stable lighting, stronger character consistency, and higher first-pass quality at 1080p. And because it lives inside Picsart, every generated video is immediately editable: refine, layer, adjust, publish — all without leaving the platform.