I always appreciate a chance to play with AI video generators. Even when they are terrible, they can be entertaining, and when they remove it, they can be incredible. So, I wanted to play with the new Runway Gen-4 model.
The company boasted that the Gen-4 (and its smaller and faster brother model, Gen-4 Turbo) can surpass the previous model of the Gen-3 in quality and consistency. Gen-4 nails supposed to the idea that the characters can and should be alike between the scenes, as well as a more fluid movement and an improved environmental physics.
It is also supposed to be remarkably good to follow the instructions. You give him a visual reference and a descriptive text, and he produces a video that looks like what you imagined. In fact, it really looked like the promotion of his own creator of video AI, Sora.
Although the videos that Sora is generally magnificent, they are sometimes not reliable of quality. A scene can be perfect, and the next one could have floating characters like ghosts or doors leading to nowhere.
Magic film
Runway Gen-4 presented itself as video magic, so I decided to test it in this spirit and see if I could make videos telling the story of a sorcerer. I prepared some ideas for a fantastic little trilogy with a wandering wizard. I wanted the sorcerer to meet an elf princess and then pursue her through magic portals. Then, when he meets her again, she is disguised as a magic animal, and he transforms it into a princess.
The goal was not to create a blockbuster. I just wanted to see how far Gen-4 could stretch with a minimum entry. Having no photos of real wizards, I took advantage of the newly upgraded chatgpt image generator to create convincing fixed images. Sora may not explode Hollywood, but I cannot deny the quality of some of the images produced by Chatgpt. I made the first video, then I used the runway option to “repair” a seed so that the characters are consistent in the videos. I reconstituted the three videos in a single film below with a short break between each.
IA cinema
You can see that it is not perfect. There are movements of strange objects and the coherent looks are not perfect. Some background elements were strangely sparkled, and I would not put these clips on a theater screen for the moment. However, the real movement, expression and emotion of the characters were surprisingly real.
And, I liked the iteration options, which did not overwhelm me with too many manual options, but also gave me enough control so that I had the impression of being actively involved in the creation and not only to press a button and to pray for consistency.
Now will he delete the many partners from professional filmmakers from Openai? No, certainly not for the moment. But I would probably experience it at least if I was an amateur filmmaker who wanted a relatively cheap way to see what some of my ideas might look like. At least, before spending a ton of money for the necessary people so that the films are as powerful as my vision for a film.
And if I became enough comfortable with him and well enough to use and manipulate AI to get what I wanted every time, I don’t even think of using Sora. You don’t need to be a wizard to see that the spell track hopes to launch on its potential user basis.
