The avatar was good. That was the problem. I'd spent maybe forty minutes inside HeyGen on a Tuesday night — kids down, laptop open on the kitchen table, reheated coffee going cold again — and the AI version of me looked presentable enough that I kept going when I should have stopped and read a single help article. The avatar nodded. The lips synced. I felt like a guy who had figured something out.

Then I exported and the whole thing was 16:9 widescreen. A horizontal bar of content, letterboxed on both sides when I dropped it into a Reel, my AI self looking like he was delivering a deposition in a very small courtroom. I posted it anyway, which tells you everything about my judgment at 11 p.m.

The HeyGen aspect ratio setting is not hidden. It is, in fact, right there. I just did not see it because I was excited, and excitement is how I waste an evening.

How a Free Trial and One Decent Reel Got Me Into This

A few weeks earlier I'd been poking at HeyGen just to see what it did — the kind of low-stakes experiment I run when something is compiling and I've already burned through my survey queue. That session went well enough that I wrote up what happened when I let HeyGen make my Reels for a week. Short version: the tool worked, the results were mixed, and I posted some clips that went nowhere but one that went a little further than I expected.

That one clip was the problem. One decent number on a Reel and I started treating myself like someone with a content strategy. I decided I was going to make a short series — same avatar, same vibe, consistent posting. I had a whiteboard plan. The whiteboard plan lasted about as long as my Duolingo streaks usually do.

My brother-in-law had been texting me about HeyGen specifically, which I should have taken as a warning sign. He'd seen something about it online and wanted to know if it was "the future of personal branding."

"You could just clone yourself and let the clone do the posting while you sleep," he said, completely serious.

I told him I'd look into it. I did not tell him I'd already been using it for two weeks and had posted exactly three videos with a combined engagement of about what you'd expect from a guy no one is looking for. He didn't need to know that part.

The Aspect Ratio Thing, and Then the Voice Cloning Thing

Here's what actually happened over about a week and a half of trying to make this work properly.

The aspect ratio issue was first. HeyGen lets you set the canvas to 9:16 for vertical video — which is what you need for Reels, Shorts, TikTok, anything a phone watches — but the default when I was building was wider. I didn't change it. I exported four videos before I figured out why they all looked wrong on mobile. Four. I remade them. That's maybe three hours I'm not getting back, and I made about the same nothing either way, so the cost was purely in dignity and Tuesday nights.

The 30-second Reel format was its own puzzle. HeyGen can generate short clips, but the sweet spot for the tool seems to be slightly longer than what performs on short-form video right now. I kept trimming scripts down, then the avatar would rush through the words and sound like someone reading terms and conditions. Getting the pacing right for a genuine 30-second reel meant rewriting the script three or four times per video. At that point I was writing and rewriting more than I would have if I'd just talked into my phone like a normal person.

Then I found the voice cloning feature and spent an embarrassing amount of time in there. You can train it on your own voice, which I did. It's genuinely impressive and also slightly unsettling — my AI self sounds enough like me that my oldest walked by the laptop and said "Dad, why are you talking about marketing?" I was not in the room. I want to be clear: I was in the other room, and my laptop was talking. That's where we are now.

Someone in a forum mentioned you could theoretically get the voice clone to sing. I tested this. The results were not good. The voice cloning in HeyGen is built for speech, not melody, and what came out sounded like a software update trying to cover a country song. I closed that tab and never spoke of it again until just now.

The 7 free days question — because I've seen people ask this — is something I genuinely cannot answer with a straight face because the trial structure seemed to change while I was using it. I don't know what the current terms are, and I'm not going to guess at them. What I can tell you is that I hit a credit wall faster than I expected, which is how these things go. I've made that mistake before with tools I got excited about, and I've written about why you shouldn't buy the gear before the first dollar — though in this case the "gear" was a subscription tier I almost upgraded to before I checked my actual numbers.

What the Letterbox Reels Actually Cost Me

Time: somewhere between eight and ten hours across about twelve days. That includes the setup, the reshoots when I figured out the aspect ratio, the voice cloning experiments, the singing thing I will not elaborate on further, and the rewriting of scripts that should have taken twenty minutes each.

Money: I stayed on the free tier longer than made sense, then paid for one month to see if having more credits would change my output in any meaningful way. It changed my output in that I made more videos. It did not change the engagement numbers in any way I could attribute to the tool versus just the content being unremarkable.

What I earned from the videos themselves: nothing directly. The Reels I made with the AI avatar performed about the same as the Reels I make from my car between DoorDash pickups. One did okay. The rest disappeared into the algorithm like a text I sent to the wrong number.

The honest accounting here is that this sits in the same column as most things I try — not a disaster, not a win. Just hours. I keep a rough mental ledger the way I keep track of promo codes: saving forty dollars is genuinely easier than earning it, and sometimes the smartest move is recognizing when you're spending ten hours to make something that earns nothing instead of spending one hour to not spend forty dollars. I did not make that smart move here. I kept going because the avatar was good and I liked watching it.

What I Kept After the Letterbox Era Ended

The voice clone is still set up. I haven't deleted it. There's something useful in there for the right project, and I haven't found that project yet, but I'm not in a hurry. It's the kind of thing that sits on my whiteboard under "figure this out later," which is where most of my ideas live permanently.

The aspect ratio thing I will never get wrong again, which is the kind of lesson that costs nothing until it costs you three hours and then it's free forever. Always set the canvas to 9:16 before you build anything. Check it before you export. Check it again. I'm putting this here because I looked for this information when I was confused and found mostly forum posts from people who were also confused, which is how the internet works.

I dropped the idea of a consistent AI-avatar series. Not because the tool doesn't work — it mostly does — but because the part that requires the most work is still the writing, and if I'm going to spend forty minutes writing a script, I might as well be the one on camera, which costs nothing and takes less time once you stop caring how you look. I care less every month.

The brother-in-law asked me last week if I was still "doing the clone thing." I told him I was taking a break to reassess. He said that sounded like giving up. I said it sounded like not paying for a subscription I wasn't using, which is the kind of boring part of side hustles that nobody puts in the video. He changed the subject to a new opportunity he'd heard about involving storage units.

The avatar is still in there, saved, looking calm and professional in a way I have never managed in real life. Maybe I'll find the right use for it. Maybe I'll log back in someday and the trial will have reset and I'll spend another Tuesday night making letterbox videos by accident. Both feel equally likely from here.

Note: Everyday Hustle Jeff is the written side of a digital series. These accounts are composite and dramatized. Any figures are what they were on the day, for one person, and are not a projection, a track record, or advice.