Is Using AI to Create Videos Good or Evil? In the author's column for AIN.UA, the Creative Lead of the agency and action, Ivan Ishchenko, using the example of commercials for such brands as Comfy, Lamel, IT Arena, and Galychyna, tells why to do this.

About a year ago, I saw the animated short Leaving Home by Erik Winkowski. And I immediately wanted to find out how it was made. It turned out that the video was created in Stable Diffusion. And it was "Wow, I want to do that!".

A year ago our team was thinking about new opportunities for our agency and the ways to optimize costs and time, and AI could help. It was time of dynamically changes: Midjourney began to make hyperrealistic images, and Stable Diffusion also received updates. I was waiting for access to Runway's Gen-1 and Gen-2 demos that helped convert text or images to video. And he gradually added his experiments to the portfolio.

Comfy case

My first attempts at contact with AI were crooked. About half a year passed until I could show something substantial at a presentation to our partners, banda, and within a few days, they came to us with a request for a video clip for Comfy. The style of the references and what I learned to do with Stable Diffusion matched.

I was thrilled! It was more than just my experiments in R&D. It was finally a commercial request for a big brand.

"When I curiously asked our Animation Lead Vitalii Nebelskyi how long it would take for an illustrator to create such a picture that I had already generated in Midjourney in a few minutes, he said three days plus a day for editing. Then I asked our illustrator how much she would do it. In general, she said that she would not take this job because it is too difficult. "

In general, we generated all the illustrations by AI in these two videos but refined the details. For example, there is a mermaid man in the video (on the cover). I developed it not only in one button press by AI. Firstly, I created a collage: the tail, the body of an antique statue, and the face and hands separately. Then I downloaded it to Stable Diffusion - and he released his version. Everything was pretty cute except for the face and hands. So I went to Midjourney, generated portraits of baroque beauties with perfect skin, and then collaged them with a mermaid body in Photoshop. And the hand was already finished by the illustrator. In the video, the mermaid and other characters speak, and we also hand-painted these micro-pictures of the mouth.

AI makes the base, and the talent adjusts all the details.

Lamel Cosmetics

It is also interesting to use AI not only to save resources but also to make the video visually more enjoyable for the viewer in general. In the video for Lamel Cosmetics, we used AI in the part where the model is supposedly walking through the desert. We generated this background using AI instead of painting each frame by hand.

It was a spontaneous idea. We liked the video, but we wanted to emphasize the durability of the eyebrow gel fixation. Classically, when something like this is planned, it is needed to shoot on the greenscreen, cut out, and the background is drawn. It would have taken a week in post-production. Now, with DaVinci Resolve and AI, we cut out the non-green background in five minutes and added the background animation generated by Stable Diffusion in half an hour.

IT Arena

For the video for the opening of the IT Arena Conference, we wanted to use AI to the maximum - a tribute to IT. The grand opening was in the Opera House, so we added a video of the symbolism and made several parts of the performance.

First, something metaphorical was generated in Stable Diffusion to lead the audience to the keynote. It looked artistic and combined technology (IT) and art (opera) incredibly.

To discuss the future, we used Midjourney and GEN-2, where we visualized flying cars, quantum computing, and AR for all areas of life. For the first time, we made the host herself not manually but in a program for working with 3D and using AI tools.

We liked the result of Stable Diffusion, so we made a fan video listing project team as actors in the intro of the Netflix series.

Can AI completely replace humans in video creation? I don't think so. It is just an additional working tool for us. If we write a standard prompt in Midjourney, then the result will be typical, like from the drains. AI will not think about what this video should contain and which audience it will work for and will not put another "Easter egg."

I think asking: "How much would AI generate this video?" is wrong. No matter how much, because he won't do it himself, it will be a completely different video category.

A lot of content is being produced now, and among this creative ocean, you need to stand out and be remembered by the viewer. Creativity regarding ideas should come from the team, and AI helps to generate these ideas in an extraordinary form.

Tisno

I currently see massive potential in artificial intelligence. It is not only about saving time and money. Suppose your customer is ready for visual experiments. In that case, AI offers vast opportunities, which applies to large businesses and a small family bakery, "Tisno."

For "Tisno," we made several promo videos with the help of AI. For example, we were inspired by the opening credits of the series "White Lotus."

This video got a lot of views, but we spent little time on it because AI generated both the illustrations and the animation. The video was inspired by the location of the establishment and its pastries. And, of course, AI-generated bugs: for example, in the video, the shrimp had two tails instead of a head, and the salt on the brownie turned into flowers. And if I didn't know how to fix the first bug, then the result with brownies is unexpectedly artistic.

As you can see, AI does a lot, but at the same time, there is still a lot of human intervention. Conventionally speaking, the robot vacuum cleaner cleans the apartment well, but it can stumble on the threshold and scatter garbage. And then you clean it yourself. Human editing is indispensable.

You can stick to Midjourney or Stable Diffusion as much as you want, but you're wasting your time if you need help understanding what it's for. For me, AI is just another tool at work, and it's worth understanding what it will bring to your team.

And when you know, sit down on the weekend and try to use at least one of these tools. You will like the result.