I spent two years making one thumbnail per video because a second one cost another hour

I was a creator before I was a founder, and my worst hour of the week was always the same: 1am, video already exported, still dragging text boxes around in Canva trying to make something that didn't look homemade.

The design work wasn't what frustrated me. It was that I had no way to test. One thumbnail was expensive enough in time. Five to compare was out of the question. So every upload was a guess, and when a video underperformed I usually knew why and could do nothing about it.

I built Miniagen to remove that cost. You give it a video link, a written idea, or an existing thumbnail, and it returns several genuinely different directions you can refine and export in minutes. Saved avatars keep your face consistent so the channel still looks like the channel.

Where it is now: around 1,200 creators, roughly $1K MRR. The feedback that keeps coming back is the exact thing I wanted for myself, which is that people finally test hooks instead of guessing.

What I got wrong early on: I optimised for output quality when the real unlock was output variety. My first version produced one good thumbnail. Nobody cared. The moment I made it produce five different directions, retention changed.

Two open questions I'd genuinely like input on. First, creator tools have brutal churn because usage is tied to publishing cadence and cadence is inconsistent. I'm running a lifetime deal partly to build a base that isn't churn-exposed, but I'm aware that trades future revenue for present cash. Second, agencies are my best-retaining segment and I haven't built anything specifically for them yet.

Site: miniagen.co
The lifetime deal, if it's useful to anyone here: earlybird.so/products/miniagen