Link to video down below 👇Stream in 1080p or Download for 1440p quality.
I get overly excited when the AI video world receives an upgrade, and a couple days ago we received a HUGE one! I have been playing around with this model since the hour it was released (that sounds sad🤔) and I gotta say, I am SUPER impressed and really excited about what the near-future holds! This video is just a string of clips to music, no extra editing has been done.
https://drive.google.com/file/d/1Ll25SfwY0f64JPmTgi-cjceU3YuuHu2U/view?usp=sharing
Here is a compilation of some of my initial clips generated using Wan2.2 compiled into a music video. Generated in HD 1280x720 and upscaled to QHD 2560x1440.
The body contact is far superior IMO and the background ambience, camera dynamics, body physics and prompt adherence are seriously blowing my mind (and other things). The facial expressions are the best I have seen. The Clothing! The detail of the clothing is insane compared to previous models. This model is an absolute BEAST and I really look forward to working with it more.
Another piece of great news is that so far with my custom workflow, the Loras (custom models) from Wan2.1 seem to be working (not all, but most). This is a huge bonus as I rely on custom models for a lot of my content creation. However, I am finding through my testing that this model is so freaking good, you don't always need one.
The only downside so far is generation time, but I have already found ways to shorten the time it takes and of course the open-source community is working hard to provide optimizations for the model.
Some of these clips dropped my jaw. Let me know what you think!
B. G.
2025-08-08 10:31:28 +0000 UTCBert lundquist
2025-07-31 11:29:33 +0000 UTC