Innovate Futures @ Benji

Wan 2.2 Img2Vid Testing - Can Wan 2 2 Run Long-Length Video Yet?

Added 2025-08-08 13:15:59 +0000 UTC

In this video, we dive deep into WAN 2.2’s image-to-video capabilities to answer a key question: can WAN 2.2 handle long-length video generation effectively yet? Through hands-on testing, we explore the first-last frames to video method, examine how multiple frames can be used as starting points, and push the limits of video duration to see when quality starts to degrade. You'll see real examples comparing WAN 2.2 with the older WAN 2.1 VAE, highlighting differences in smoothness, color consistency, motion quality, and prompt adherence. We also test looping workflows using travel prompts, frame overlaps, and for-loops to extend video length, revealing practical tips and limitations. This content is ideal for AI video creators, generative model enthusiasts, and developers working with diffusion-based video generation who want to understand the current strengths and weaknesses of WAN 2.2. It matters because choosing the right model and method can save time, improve output quality, and help you avoid common pitfalls when generating extended AI videos.

Attached Workflow to testing the multi-frames for FLF2V workflow.