The biggest bottleneck in D2C paid ads isn't creative quality. It's volume. You can't predict which ad wins until you ship it, and most teams can only afford to test one or two formats at a time. So we built a workflow on Agent One that brings the cost down to $75 per creative and under an hour per ad. Not basic AI clips. Full UGC ads with hooks, A-rolls, B-rolls, voiceovers, and green screen composites. In this video we walk through the entire workflow from start to finish, and break down exactly what each ad costs and how long it took to produce. Watch the full tutorial to start shipping winning ads.

So we created our own character sheet and uploaded it to Agent One. We tested it across multiple different shots, and it worked every single time. That’s it. It’s that simple. You just need to give the model exactly what it needs; don’t overload it and confuse it. If there's something specific you need to lock like a ring on his finger, a clip in her hair, or a chain she's always wearing, add a small box for it on the sheet. Just that, and it holds across every shot.

To the naked eye, the animated shot looked right. But when you zoom in, the details start to slip. The skin texture isn't quite the same. The jawline shifts slightly frame to frame. The hair colour drifts. If you're doing real commercial work, these minute details matter. Clients will catch every one of them in review, and that's exactly what kills the project.

So we uploaded it to Agent One and animated it. The face that came out barely looked like the one we put in. Sheets like this look impressive, but the model doesn't actually read most of what's on them. Storyboards, camera maps, cinematography notes, none of that helps it hold a face. It just adds noise.

Every week there's a new "character consistency" workflow going viral on X, and 99% of them don't hold when the camera actually moves. Hridaye, our creative director, has been an AI filmmaker for 3 years and has tested literally every single one of them. So he pulled up the two most viral character sheets making the rounds right now, tested both, and then shared the one that actually holds every single time. Character consistency isn't just a preference when you're making AI films. Clients are sticklers about it, and if one face doesn't hold across shots, the whole project's dead. Watch the results for yourself in the video and the thread.

If you run paid ads for a D2C brand, you already know the problem - you spot a winner in the ad library and recreating it from scratch is weeks of production and a tonne of money. So we built a workflow on Agent One that lets you upload a winning competitor ad and rebuild it for your own brand. The agent watches the ad and breaks it down - not just the shots, but what actually makes it a winner. Then it tells you what to keep and what to swap for your brand, builds a 9-panel storyboard, and once that's locked, hands you a full shot list with timestamps, camera moves, and composition before it starts batch-generating. Line the clips up in Slate, invideo's new feature, and you never have to leave the platform,  even for the edit. Check out the full video for the step-by-step walkthrough.

8 binge-worthy AI series that had our in-house directors completely hooked 🧵👇

#7 SAGE by @mikejmitch After years of war, Elijah Kelly returns home to Mill Ford. But he finds something far worse than what he left behind. The town is not empty. It is waiting. What's impressive about how Mike J Mitch approaches this is how he breeds the horror directly into the character design instead of layering it on as a flashy visual effect. By embedding the supernatural threat into the physical grime and wardrobe of the period, he avoids the superficial look of most AI films and makes the nightmare feel entirely native to 1860s Nevada. Watch it here: nitter.app/mikejmitch/status/2033…