Seedance 2.0: It's not just slop anymore

I have not posted a video on YouTube for months. I just haven't had any interest in it. What I have been interested in is ComfyUI and video/image generation. This has also led to interest in open source models in general. It's actually the first time in a while I've had something hold my interest. Learning something new has its drawbacks though, like sucking at it. It takes time to grasp a completely new set of concepts like Python, CLI, and Linux commands. This is just the stuff you need to operate the apps — not learning the apps themselves.

You watch the tutorials that don't tell you important things — like, I don't know, Windows being a nightmare, and dependencies breaking every update — but it's kept my interest. I made a couple of videos using subscription websites, but being Gen X, I want to own what I use, not rent it from middlemen. It's a good thing I've been building my own computers since 2000 and have workstations that can handle this new field. I've downloaded terabytes of files and seemed to be moving along when out of nowhere comes something amazing.

It's not really nowhere, but China, where most of the open source models come from. This one is closed though, and once again you have to pay a subscription fee. It's worth it though. It has been trained so well on films and videos that it can create amazingly high-quality scenes. There is no open source model that can even come close to its intuitive understanding of cinema. Camera movement it generates on its own, sound effects and dialogue — the whole works. Right now it's being used to create fights and fantasy clips using Tom Cruise and other icons. I'm sure this will not go on much longer, though, as it is giving Hollywood studios a heart attack.

It's called Seedance 2.0, and it is literally a "game changer" — a term usually assigned to the unworthy. In this case, though, it is legit. The untrained can make good clips, and the trained? Well, we will see.

Now, back to ComfyUI.