WaveSpeedAI

Wan 2.7 Is Coming: A Major All-Around Upgrade Over 2.6

wan alibaba video-generation
Wan 2.7 Is Coming: A Major All-Around Upgrade Over 2.6

Wan 2.7 Is Coming Soon

The next generation of Alibaba’s Wan video model is almost here. Wan 2.7 is planned to launch within March 2026, and it’s shaping up to be the biggest upgrade the Wan series has seen yet — a major all-around leap over Wan 2.6.

If you’ve been following the Wan model family, you know that each release has pushed the boundaries of open-source AI video generation. Wan 2.7 is no exception. This time, the improvements span nearly every dimension of video creation.

What’s Improving in Wan 2.7

Wan 2.7 brings significant improvements across five core areas:

  • Visual quality — Expect sharper, more detailed, and more photorealistic outputs with better color accuracy and fine-grained detail preservation.
  • Audio — Enhanced audio generation capabilities for more natural and synchronized sound in generated videos.
  • Motion dynamics — Smoother, more physically plausible motion with better temporal consistency across frames.
  • Stylization — A broader and more controllable range of artistic styles, from cinematic looks to animation and beyond.
  • Consistency — Improved character, scene, and object consistency throughout the entire video, even in complex multi-shot scenarios.

These aren’t minor tweaks — they represent a meaningful step forward in what open-source video models can deliver.

New Features in Wan 2.7

Beyond the quality improvements, Wan 2.7 introduces several powerful new capabilities that expand what’s possible in AI video creation:

First-Frame & Last-Frame Video Generation

Specify both the starting and ending frames of your video, and Wan 2.7 will generate the motion in between. This gives creators precise control over the narrative arc of each clip.

9-Grid Image-to-Video

A novel approach to image-to-video generation that uses a 9-grid layout for richer, more structured input. This enables more detailed scene composition and multi-angle reference for higher-quality video output.

Subject + Voice Reference

Combine a visual subject reference with a voice reference to generate videos where both the appearance and voice of the character are consistent with your inputs. This is a game-changer for personalized content creation.

Instruction-Based Video Editing

Edit existing videos using natural language instructions. Want to change the background, modify lighting, or alter a character’s outfit? Just describe the change and let the model handle the rest.

Video Recreation / Replication

Recreate or replicate existing videos with modifications — whether that’s changing the style, swapping subjects, or adapting content for different contexts while preserving the original motion and structure.

A More Powerful Creative Workflow

Together, these improvements and new features point to a more comprehensive creative workflow. Wan 2.7 isn’t just a better video generator — it’s evolving into a full video creation and editing toolkit.

With first-frame/last-frame control, instruction-based editing, and video recreation all in one model, creators will be able to go from concept to finished video with far less friction than before.

Try Wan 2.5 & 2.6 on WaveSpeedAI Right Now

While we wait for Wan 2.7 to drop, you don’t have to sit idle. Wan 2.5 and Wan 2.6 are already live on WaveSpeedAI with a full suite of capabilities ready to use today.

Explore the Wan 2.6 collection →

Wan 2.6 offers text-to-video, image-to-video, reference-to-video, image editing, video extend, and more — including Pro, Flash, and Spicy variants for different speed and quality needs.

Explore the Wan 2.5 collection →

Wan 2.5 remains a rock-solid choice for text-to-video, image-to-video, text-to-image, image editing, and video extend workflows.

All models are available through WaveSpeedAI’s API with no cold starts and competitive pricing. Getting started takes just a few lines of code:

import wavespeed

output = wavespeed.run(
    "alibaba/wan-2.6/text-to-video",
    {"prompt": "A cinematic shot of a wolf running through a snowy forest at golden hour, with snowflakes catching the warm light"},
)

print(output["outputs"][0])

Get Ready for Wan 2.7

There are still two weeks left in March, and the possibilities are wide open. Wan 2.7 is set to be a landmark release for the open-source AI video community.

In the meantime, start building with Wan 2.5 and Wan 2.6 on WaveSpeedAI — everything you create and learn now will carry forward when Wan 2.7 arrives. Stay tuned for the launch announcement.