OK, the title is doing a lot of work. Let me knock it down to size before you decide I'm selling you something.
Last month my quit-vaping app cleared $6,200. That's the number, and here's where it comes from: about $5,400 in app subscriptions, and roughly $800 from a small pack of video prompts I sell on the side. That's it. That's the income. I got there with an AI side hustle that needs no audience, no camera, and no face, which is the part I still find a little strange. I'm Callum, 27, and until last spring I was the assistant manager at a gym in Sacramento that closed when the lease ran out.
You probably clicked because of the 30 million views. That video is real. It also made me almost nothing, about $3,800 total, because I was messing around and never actually showed the app in it. So please don't read this as "go viral and get rich." A 30 million view video paid me less than three good weeks of boring subscription growth. Views are not money. I learned that the expensive way.
This is not the only AI side hustle worth trying. It's just the one with the lowest barrier I've found, because it needs no audience to start. You post into the void and the algorithm decides, not your follower count.
It also did not work at first. My first month I posted three videos a day for thirty days straight, about ninety videos, and the whole month made something like eleven dollars. Most got a few hundred views. Some got twelve. I nearly stopped around day nineteen, after a video I spent four hours on got seven views and one comment telling me the hands looked fake. The hands did look fake.
One thing before I get into the workflow. I spent about six weeks stitching together free YouTube tutorials on AI video, and every one pushed a different tool and a different "only way" to do it. I had twenty browser tabs open and not one finished video. A structured course cut through that so cleanly it was almost embarrassing how fast I got my first real clip out once someone put the steps in order. YouTube is fine once you already know what you are looking for. It is a maze when you do not.
One more thing before the steps. The single most important part of this is the last section, the one about ideas, and it is the least technical thing in here. I'm saving it for the end on purpose. Don't skim past it, because the tools are the easy part and that section is the whole game.
The tools that finally replaced my twenty subscriptions
For a while I was paying for something like twenty different AI tools and canceling half of them every month. Now it's two. For images I use Nano Banana Pro, which right now just beats everything else I've tried, and it isn't close. For video I use Higgsfield because it bundles most of the video models in one place instead of a separate subscription for each one. That's the whole stack. I'm not sponsored by either, and both will probably be old news in six months, because this stuff changes constantly. Whatever is best when you read this, use that instead.
How I actually make one video
My process is boring, which is the point. I start on Pinterest and find a reference image, usually a "before" look and an "after" look for a character. I save a small library of them. Then in Higgsfield I keep one prompt sitting there permanently, something like "replace the second person with the first person." I drop in my reference, pick Nano Banana Pro at 9 by 16, and generate a few options.
Once I have an image I like, I turn it into video. I don't do big multi-shot scenes. I add one small movement, like "he slowly looks up and exhales." Slow beats fast every single time, because fast movement is where the model falls apart and you get melting hands. I run it on Kling 3.0, or at least 2.6 if I want to save credits. And I generate at 720p on purpose. Higher resolution looks too clean. 720p matches the slightly rough iPhone look people expect from a real phone, so it reads as more real, not less.
The trick that made it feel like me
The move that changed everything was motion control. I film myself doing something on my actual phone, a shrug, a walk to the fridge, pouring a coffee, and then I lay the AI character over my own footage. The timing is mine, the body language is mine. It's basically me wearing a different face. That's what fixed the fake feeling. When the motion comes from a real person, the uncanny edge mostly goes away.
I learned the prompt side of this from a mix of YouTube and a short course on Mindwand that broke AI video into fifteen-minute daily lessons I actually finished. The daily format is the only reason I stuck with it. Skool and Coursera have similar material if those fit you better. Pick one. Finish it.
Editing and posting without overthinking it
I pull the clips into CapCut, or honestly I just edit straight inside TikTok, which is faster because it's native to where I'm posting. I keep constant motion on screen, drop in a trending audio, plug the app once, and post. Hashtags barely matter. What matters is whether the video makes someone stop and comment. If your video leaves people totally indifferent and they scroll on with no reason to even check the comments, you failed, and no amount of hashtags saves it.
And you have to keep posting. Three a day for at least thirty days before you judge whether any of it is working. I see people quit on day six all the time. Thirty days is the floor, not the finish line. I got a little lucky and had one video pop early, but most of the time it's slow and quiet for weeks.
The part I saved for last: the idea is the whole thing
Everything above you can learn in a weekend. The tools don't matter. The idea behind the video is the entire game, and it's where almost everyone loses.
Here's how I come up with them. I don't just scroll for inspiration, that's my worst method. I read, I look at random images and paintings, and I write down as many ideas as I can, going as extreme as I can before pulling back. Two things guide the pullback. First, aim at a real pain point. The deeper the pain the video speaks to, the harder it lands, so I think in terms of Maslow's hierarchy and try to hit something low and basic, not surface level. Second, an old David Ogilvy idea: sell the toothpaste that makes teeth whiter, not the one that prevents cavities. People buy the dream outcome, not the problem you're removing. So I show what someone's life looks like after they quit, and I barely mention the app at all. Think of the old Apple ads, or Red Bull, which is basically a marketing company that happens to sell a drink. They sell the feeling. The product just shows up nearby.
The niche pull is real, for what it's worth. Most people who vape say they want to stop, according to the CDC, so the pain point isn't something I had to manufacture. It's already sitting there.
That 30 million view video? It was the most out-of-shape guy imaginable, then a caption about quitting a bad habit, then a completely different, transformed AI person a year later. It hit a real pain point and it got people arguing in the comments. It also barely converted, because I got the marketing right and forgot the boring part, which is actually pointing people to the thing you sell. I fixed that on the next ones. Now a video with a hundred thousand views out-earns that 30 million view one, because those hundred thousand people actually get told where the app is.
If you want the prompt and character side to click faster, I picked the prompting course on Mindwand because the daily-lesson format fit around my old gym shifts. Could have used Skool or Coursera, it honestly doesn't matter which. Format matters more than brand, so just pick the one you'll actually finish.
So what's the real hourly rate. If I average every hour, including the ninety near-worthless videos from month one and all the failed generations, I'm at maybe sixteen dollars an hour over the life of this. Good months feel like way more than that. The month I made eleven dollars felt like a lot less. That's the honest version the "I made 7k on the side" videos leave out.
I'm not going to tell you to quit your job. I'd tell you to pick one bad habit a lot of people share, film thirty seconds of yourself moving, put an AI face over it, and post it. See how it feels when your first one gets seven views. If that doesn't put you off, you might actually last the thirty days, which is more than most people manage.