Wan 2.7 Preview: Full Upgrade in Quality, Audio, Dynamics & Instruction Editing
Alibaba's Wan 2.7 launching March 2026 with comprehensive upgrades. Multi-image grid to video, instruction-based editing, real person input, subject + voice reference.
Wan AI Team
Wan AI

We're thrilled to preview Wan 2.7, launching in March 2026 — one of the most significant upgrades in the Wan series.
Comprehensive Quality Upgrade: Wan 2.7 delivers full improvements across image quality, audio, dynamics, stylization, and consistency compared to Wan 2.6. Every aspect of video generation has been enhanced.
Multi-Image Grid (9-Grid) to Video: A revolutionary new feature allows you to generate videos from a 9-grid layout of images, enabling precise multi-scene composition and storyboarding.
Instruction-Based Video Editing & Replication: Edit existing videos using natural language instructions, or replicate a video's style and motion. This opens up entirely new creative workflows.
Subject + Voice Reference: Provide reference images for character appearance and audio samples for voice tone. Wan 2.7 will generate videos with matching subject identity and voice characteristics.
First/Last Frame Generation: Specify the starting and ending frames of your video for precise control over the narrative arc and visual transitions.
Real Person Image Input: Unlike SeedDance 2.0, Wan 2.7 supports real person photographs as input, with up to 5 video reference clips and flexible 2-15 second dynamic duration at 1080P.
Stay tuned for the official launch in March 2026!


