Creating Product and Creator Assets

00:00 Creating realistic UGC ads normally means paying a creator for every single video, shipping them your product, and praying that what comes back is usable. But I found the best workflow to make your own professional UGC ads using nothing but AI. And in this video, I'll show you the whole process. So, by the end, you can make your own highquality UGC ad in different languages, reaching multiple markets at once. The first step

00:22 is to actually create an image of our product, which becomes the foundation for everything that follows. The product I'm using is an energy drink called Pop Wire, and it's completely made up, so there's no photo of it anywhere and nothing to shoot. That's on purpose since if this holds up on a product that doesn't exist, it'll hold up on whatever you're actually selling. The tool I'm using for everything in this video is called Open Art, which is an all-in-one

00:45 platform that gives you every AI model you need in one place. If you want to follow along, I've left a link to it in the description below. When you first log in, click on tools and you'll see the main navigation bar with video, image, audio, character, and world on it. We'll start in image. Click on that, then click create image, and it takes you straight into the workspace. From the model selector, I'll go with GPT image 2.5 and set the mode to sunburst

01:08 because it's the strongest model here for making readable text onto an image, which is very important when you design a product. Then I'll push the resolution up to 4K. Set the aspect ratio to 16x9. Put the quality on high and turn autopolish off. Now I'll paste in my prompt. And you can find this prompt along with the entire workflow down in the description completely for free. So you can follow along in real time. What

01:31 it builds is one sheet with 10 panels of the same can. So the model can see exactly how it looks like from all angles when we animate it later on. And I'm also spelling the brand name out letter by letter in the prompt instead of just typing it once. That way we ensure the text comes out accurate instead of glitched, which is something image models tend to do. Right after that, I make it clear nothing else appears on the can. That's what stops

01:54 the model from tacking on extras I never asked for. So, let's generate. And all 10 panels come back matching. It's the exact same can in every single one with every angle we could need right down to the pore. The close-up is the one I check first, though, because the name and the flavor line are sharp enough to read, so the label is actually accurate. So, every angle and every state of this can is now in one image and nothing

02:16 after this has to describe it again. But nobody buys an energy drink because of a catalog photo. We need an actual creator to promote it. And that part is the hardest to get right. And they need to be both engaging enough to promote the product and casual enough that the final video doesn't read like an ad, which is what UGC should feel like. I'm staying in the image generation workspace with GPT image 2.5 on Sunburst, and every setting stays the same. I'll paste in

02:40 the next prompt. What this one builds is a character sheet, which is one image split into three panels with the same person and the same outfit in all three. On the left is the full body with no head. In the middle is the full back view, and on the right is a tight closeup of just the face. That left panel is intentional since it's meant to lock down the body and outfit, while the right panel handles facial consistency. Most of the prompt goes into describing

03:03 his details and imperfections like the moles on his cheek and the small [music] scar, which is what makes him look both unique and realistic. It also asks for him to look like an ordinary, believable person instead of a model. That matters more here than in any other kind of video because a UGC ad is supposed to look like a normal person filmed it. And a face that's too perfect is the first thing that gives it away. The last part

03:25 is the lighting, which asks for completely flat light with no shadows anywhere on one plain gray background. Any lighting that's already in this image gets carried into the video later, and it won't match the room we put him in. So, his face ends up looking different from shot to shot. I'll generate, and all three panels come back as the same guy with the same face, the same outfit, and the same flat gray background. The skin is the part that

03:48 sells it because there's real texture in it instead of that smooth plastic look that gives it away as generated. So that's our creator done and he'll come out identical in every ad we put him in from now on. Now that both our assets are ready, we need to actually start making an ad. And this is the one generation that has to do everything at once. The acting, the product, the voice, and the cuts all in a single pass. So back to the navigation bar, and