Stop Consuming, Start Creating

I woke up this morning to LTX-2 model announcements. It is a new Open Source model that generates video AND audio, with built in control net support (poses, edges, depth) and more. “Wan 2.2 is dead!” the video titles proudly announce.

With AI image and video generation there is rapid advancement going on. New and better models are almost constant. And they frequently are better. If they were not, it would make things easier.

A challenge I find for myself is “should I learn this new model first, or keep going with what I know already?”

Two contrary pieces of advice:

• If the old technology is enough, stick with it. It takes time to learn new technologies. You have to discover all its pros and cons, how to work around its limitations, etc.

• Be aware of new technologies. They frequently do solve problems and make you more efficient. It can make your more efficient in creating content, and changing technologies can be visibly different in the results you create, so if creating a series, people can notice the difference.

The challenge is to decide between them. If you goal is to create content (which is my personal goal), then sometimes its better to stick with the old technology and create. Get it done. Once done, then look at the new shiny technology.

Where I continue to struggle personally however is when I have not created a final result yet, and the current technology is painful to achieve a result. Here I look closely at the new technology and decide whether to move or not. For LTX-2, it sounds very promising. But there are limitations. For example, I don’t know how to make a voice consistent across scenes. So I may tinker with LTX-2 to see if it solves problems for me that other models do not, but I am trying to do so from a perspective of “I don’t want to change my pipeline - I just want to replace parts of it.” If the new technology does not fit in, then move on no matter how cool it seems.

But I won’t pretend its not hard to make these decisions.