How I make AI UGC reels for about three dollars each
The subscription that gave me two videos
I run a side project, and like most people building something on their own, I wanted to advertise it on social media without being on camera myself. So I went looking for a way to make faceless videos.
I signed up for one of the commercial tools and paid for a monthly token subscription, twenty-five dollars, with no real sense of what those tokens actually bought me. I started building. It became clear fast that tokens disappear quickly when you're trying to get a specific video out of your head rather than whatever the tool feels like giving you. A word gets mispronounced, a background changes between shots, and there's no way to fix just that. You rebuild the whole thing from the start, tokens and all. That subscription gave me about two videos I was happy to post. I wanted to post at least three times a week. The maths never worked.
The DIY route that cost more, not less
So I went back to research, this time on YouTube, where you can find no end of creators showing you how to build your own faceless videos by connecting Claude or ChatGPT to a generation tool. I followed the instructions and started again. What those videos don't show you is that they're never the first attempt. You can genuinely build a reel this way, but you hit the same wall as the commercial tools: a mis-spoken word, a background that drifts from shot to shot, no real story structure holding it together. This route took longer and cost me fifty-two dollars to land one reel I liked. I was ready to give up.
The idea that actually fixed it
The thing both routes had in common was the unit of failure. Every tool I'd tried treated the whole clip as one throw of the dice: get it right, or pay again to redo all of it. My reels were short, around thirty seconds, which sounds like nothing until you realise how much can go wrong inside thirty seconds of AI-generated video.
So the question became: what if a mistake only ever cost the one piece it was actually in? Not the whole reel, not the whole render, just the clip. Three months and a lot of late nights later, that idea was OneTake.
How I actually build a reel now
The workflow is the same one I use for every reel I post:
- Cast, location and props are already built. I pick a format, an actor, a location and whatever props the shot needs. None of that costs anything.
- The script is mine to write or hand off. Typing it myself costs nothing. If I'm out of ideas, a script is written for me for 20 cents, and I can tweak it until it reads right.
- The first clip builds as a preview, not the whole reel. Hitting go only renders the first four seconds. That's the point where I decide if the look and feel is right before spending anything on the rest.
- I approve or rebuild clip by clip. Every clip after that gets the same treatment: approve it, or rebuild just that one. Nothing else in the reel is touched or re-charged.
- Once every clip is approved, the reel assembles itself. Gaps get trimmed, captions and opening or closing banners get added, a cover image is generated, and it's ready to download.
What it actually costs
OneTake charges per element, not per attempt at a whole video: $0.40 for a clip or an anchor, $0.15 for a generated image, $0.20 if I want a script written for me. A typical six-clip reel with one anchor and a written script comes to about $3.00. If one clip is wrong, fixing it costs the price of that one clip, not the reel.
That's the whole difference. Not a cheaper version of the same gamble, a system where a mistake stops being expensive.