Loading
November 22, 2026

There’s a new AI model coming out every single week, and every week there’s a video telling you that this one is finally the one that makes you money. I have saved a lot of these videos and I’m sure you have too, but you don’t take any action, and the people who actually do are already making money from it. I sell videos to make money and I also run a video editing agency, so I’ll just open the marketplace and show you what AI videos people are actually selling right now, and we’ll recreate them step by step, and I’ll also tell you exactly hohttps://drive.google.com/file/d/10LtYFwGFtE0xSab-zpMQ3QbzUO1Zav5C/view?usp=drivesdkw much time and tokens it took to generate all of these. Before we start, let me show you what people are selling right now. There are two very famous marketplaces, Fiverr and Upwork. I’m going to Fiverr right now, and here we’ll just search for AI videos or AI product videos, and you’ll see that there are close to 64,000 results, which means a lot of people are selling these services, but this is not a good metric to judge. A good metric is the number of reviews that they have, as reviews on platforms like these are only by buyers who have done business with them or purchased some service. A good metric is to see how many reviews they have and how many active orders they have. This guy has two active orders; he’s charging $85 for a 10-second high-quality AI-generated video and his price goes up to $640 for a 2-minute video, so we’ll see his demo video and I’ll tell you how to recreate it step by step. He made an AI commercial for Lay’s which looks pretty decent—you can still tell it’s AI, but it doesn’t matter because clients here want AI videos which are cheaper and they can get a lot of these. He’s giving a ready-to-use video, so let’s see how we can do that: he’s getting product photos, brand colors, and a short script or key message from the client.

Now to get started, we’ll go to OpenArt, which is basically an AI studio where you can generate these videos, and here we’ll go to video mode. We’ll specifically use Wan 3.0 because that’s the cheapest model which can generate videos up to 30 seconds. We are doing this because I want to find a cost-effective way for you to make videos and also sell them while not burning a lot of credits. We’ll select Wan 3.0 here, and you have two options: you can either upload your media files or you can create your characters and world and describe details here. What I personally like to do is also open an LLM like Claude or GPT, as AI models are usually really good at writing prompts. This is the method I usually like to follow: we go from our idea, then we generate the images so the model nowhere should be guessing, and from those images we finally create the video. The idea is to create a very similar chips packet ad, so for that we need a character, which is a guy, and then we need the chips packet.

To do that, we’ll come to Claude and write the idea. I’ll roughly describe the idea to Claude and then generate the images. We can take inspiration from this, or if you want to make it even easier for yourself you can directly take screenshots from here, but I don’t want to copy someone’s work ditto so I’ll use my imagination. You can see why I don’t prompt myself because this is a very rough idea, and if I give this directly to the AI model it will hallucinate, so from here we’ll create all the assets first and then create the video. I’m giving my rough idea to Claude and asking it to generate character images first by creating a character sheet. Claude has directly provided me the instructions, so I’ll go back to OpenArt, select GPT 2.5—which is like the best model right now—and just paste this. This will take 100 credits, which is okay, and I don’t directly go and upload this in OpenArt; I give this back to Claude to make sure we are okay with everything, and then I go ahead, as this works for the character. Next, I generate four or five frames depending on how many scenes I want to have. For example, the first one is on the couch, and here I can add visual references like the character sheet that we generated and the product image.

The first frame looks really good and our product looks great, but the product is not open, which is a problem we can fix quickly. Make sure while generating images you pay a lot of attention because the video will directly come out of this, so the more refining work you can do here the better it is. I’m uploading this image back to Claude and asking it to fix the minor details, and then we can generate all six of them. It has given me another prompt, and this time in references I can add the latest image as well; the main point is that you add all the references here each time to get the best output. This one looks good and can be our hero frame, and from here we can generate all the remaining scenes. I uploaded the final image and asked for the next frames and what I should attach as reference. This is a very important step that a lot of people skip and then their product or face slips, so here it asked me to attach frame one and the sheet. I’ll paste the prompt, attach image one and the sheet, and at the same time generate the remaining frames as well. Notice how it’s telling me what to attach in each generation, which is very important. Once we have everything ready, I’ll download all the photos, go back to Claude,.

upload them, and give a simple prompt saying the video should flow well from shot to shot and look like a commercial, asking for a prompt for Wan 3.0. It gives me a complete prompt along with what each image is. Then we go to video mode, select Wan 3.0, select the character sheet just in case, and paste the final prompt. Settings are very important: I want the audio on, and I have two modes which are standard and prime. Standard takes less credit, but the quality won’t be as good, so I’ll go with prime, select 25 seconds and 720p, keep audio on, and hit generate. The process is mostly very simple, though you need to generate each frame beforehand because you don’t want the model guessing and getting it wrong. Before you decide to use Wan 3.0, you should know what it is good at: first, it generates clips up to 30 seconds at 1080p and audio is generated alongside it so you don’t have to generate it separately. The best part is it can take up to 20 references, convert a document to videos, and it’s one of the most cost-effective models taking the least number of credits. The first category was product videos, and the second category is UGC ads. In this category there are not so many gigs, but these people still have reviews, which means there’s less competition and still demand. Here’s a gig: “I will create your AI creator VSL spokesperson UGC with editing,”

starting at $100 for a 20-second video and going up to $350 for a 60-second video. The only thing you need to get right is the lip sync and the characters, so we’ll follow the same process: generate the character image in OpenArt alongside the product image, and Wan will create a complete video for us. The best thing about UGC videos is that clients usually buy a bundle of them, requesting maybe 10 to 20 videos in a single order, making this a really good opportunity if you get it right. Category three is launch and brand videos. One seller is charging $395 for a 15-second product ad and already has two active orders, with prices going up to $595. These are not very complex videos; they are very simple videos that I just showed you how to create using the same process. Category four is animated AI videos, where the pricing is crazy—it starts from $100 and goes up to $900 for a 60-second video, and this seller has 366 five-star reviews and one order in queue. These ones are even easier to make because Wan is really good with animation, so we do the same process: create a mood board and character sheet, and then create the complete final video. In my experience, the best one is AI real estate videos, which you might have seen a lot of on Instagram.

✨ IMAGE PROMPT
A upload image highly detailed, cinematic comic-book-style digital illustration of a stylish young man with voluminous, textured black hair, wearing an oversized same dress, a smartwatch, and a minimalist pendant necklace. He same pose, looking slightly sideways with a cool, relaxed expression. Dramatic dual-tone cinematic lighting, vibrant blue rim light from the left and electric cyan-blue rim light from the right. And inner shadow blue Deep black background with a subtle yellow and teal gradient glow. Bold black ink outlines, intricate cross-hatching, realistic facial details, richly textured fabric, vivid orange skin highlights, dramatic shadows, high contrast, sharp edges, premium graphic novel artwork, ultra-detailed digital painting, 4K quality, portrait, vertical composition. Background black Note: white background

This seller is charging $60 for a 30-second video and goes up to $180 for a 90-second video. Once you have your product image—which here is the house—this is very easy to make: all you do is take the house frame, create reference images, and generate the videos. Finally, I’ve been testing Wan 3.0 in OpenArt for a while now and there are a few things you can do to avoid failures: first, it struggles with complex physics and fluids, which can look unnatural so I would avoid them, and if you are generating a very complex long scene, I suggest dividing the videos into smaller sections for better output. You can try Wan 3.0 on OpenArt using my link below, as their annual Wonder plan is the most cost-effective here, and don’t forget to subscribe.

Font link

Full Project

Photo link

XML file

Song link

Leave a Reply

Your email address will not be published. Required fields are marked *