Face drift costs you three things: credits for every re-roll, hours for every mismatch, and approval rounds when a client spots the difference. Building your own character sheet moves that cost up front instead of removing it. A talent profile built to hold the face removes most of it.
What is face drift?
Face drift is when an AI-generated person changes appearance from one image to the next, even when the prompt barely changes. You get a face you like in the first scene. Then you change the location, the outfit or the light, and a different person shows up. We explain why it happens in Your Prompt Isn't the Problem. This post is about what it costs.
What does face drift cost?
- Credits. Every regeneration spends credits, and the more scenes in a set, the more chances for the face to shift.
With a single reference image, it often takes 2 to 5 tries to get a face that matches in each new scene, and every try costs credits.
With Trueth and our built-in character sheets you usually get a solid likeness on the first or second try. Across a full set, the difference adds up fast. - Time. Checking, comparing and re-running takes hours that never appear on the brief.
- Approvals. A client who sees two different people in one campaign sends it back, and the clock restarts.
It can also cost trust. In DoubleVerify's 2026 global study of 22,000 consumers, 42% said low-quality or uncanny AI advertising would negatively affect their opinion of a brand.
DoubleVerify's CEO, Mark Zagorski, said in the company's release: "AI itself is not what determines engagement. The quality of the output does." That's the point. People aren't rejecting AI. They're rejecting AI that looks off, and a face that changes from one shot to the next is exactly the kind of quality problem that gets content written off as slop.
Why isn't a character sheet the fix?
A character sheet is a set of reference images and notes you build so the AI can keep one character consistent. It can help greatly, but it has costs of its own. You gather references, test generations, fix the eyes, the jaw and the skin tone, regenerate, and repeat for every new character.
You pay in credits and hours before you've made one usable shot, and you build it again each time you need a new face.
How does a Trueth talent profile avoid face drift?
Every Trueth talent profile is anchored in that person's training images. We spent years tuning the balance of reference images and prompting until the face almost never moves. It's much more than uploading a single reference photo. It's a proven formula that locks in a likeness without the credit burn or the hours spent re-rolling.
The result is that you cast a talent once and keep them in every scene. And every talent is a real person who signed a commercial likeness release and earns a royalty each time they are used.
How can you test it yourself?
Pick one talent and create the same scene in five locations. Put the faces side by side and count your re-rolls. Then run the same test with your current workflow and compare the credits spent.
You can create your first image free at trueth.io/try.
FAQ
What is AI face drift?
Face drift is when an AI-generated person looks different from one image to the next, even with a similar prompt. It usually shows up when you change the scene, the outfit or the lighting.
How many credits does face drift waste?
It depends on your tool, but the pattern is simple. With a single reference image, expect to re-roll a few times for every new scene. With a Trueth talent, you usually get a solid likeness on the first or second try. Multiply the extra tries by the number of scenes in your set, and that's the credits you're wasting.
Do I need a character sheet to keep a face consistent?
Not with a Trueth talent. The profile is built once and reused across scenes. With other tools, a character sheet or a set of reference images is the usual route.
Is a Trueth talent a real person?
Yes. Every talent is a real person who signed a commercial likeness release and earns a royalty each time she's cast.
Sources
- DoubleVerify, Global Study: Poor-Quality AI Content Puts Brand Trust at Risk, July 29, 2026. The full report is Global Insights: Media Quality in the Age of AI.