We've spent two years marvelling at what AI can generate. Maybe it's time we talked about what it can't repeat.
Autodesk — yes, Autodesk — just revealed that their own AI-generated film had a continuity error so stubborn that no amount of prompt engineering could fix it. When the company building professional animation tools can't wrangle their own AI into frame-by-frame consistency, that tells you something important about where this technology actually stands.
What happened
An Autodesk executive disclosed that the company's internally produced AI film contained a persistent continuity error that resisted every attempt at correction through prompting. The takeaway, in their own words: AI still lacks the precision and intentional control needed for professional animation workflows. It's a remarkably candid admission from a company that's been heavily marketing AI capabilities to creative professionals.
Meanwhile, Laika — the stop-motion studio behind Coraline and Kubo and the Two Strings — is being championed as the most relevant animation house for this moment. Not in spite of its painstaking, handmade process, but because of it. In an era of AI-generated imagery, Laika's visible human craftsmanship has become a statement in itself.
Two stories. Same week. Same lesson.
The bigger picture
Here's what I think most designers are missing in the AI conversation: the hard part of design was never generating a single beautiful thing. It was making that beautiful thing consistent.
Consistent across 200 frames. Consistent across 47 pages. Consistent across a design system with 300 components and six teams touching it. Consistency isn't boring — it's the backbone of craft. And it's precisely the thing current AI tools are worst at.
Think about what continuity actually requires. It's not just "make this look the same." It's "remember the intentional decision I made three steps ago and carry it forward with purpose." That demands a mental model — an understanding of why something was done, not just what was done.
AI doesn't have that. It generates each output as if it's the first time. Every frame, every screen, every iteration is a fresh guess informed by probability, not intention. You can guide it with prompts, reference images, style tokens — but you're essentially trying to reconstruct intention from the outside, over and over again.
This matters far beyond animation. If you're a brand designer, consistency is literally your job. If you're building a design system, consistency is the product. If you're designing a multi-page experience, consistency is what makes it feel like one experience instead of twelve unrelated screens stitched together.
And this is where Laika becomes such a fascinating counterpoint. Stop-motion animation is absurdly labour-intensive. Every puppet is hand-sculpted. Every frame is manually positioned. It's the opposite of generative — it's deliberate, controlled, painstaking. And the result is something audiences can feel. There's a coherence to it, a sense that every detail exists because someone chose to put it there.
That's not nostalgia. That's craft.
Consistency isn't a feature you bolt on. It's the result of sustained, intentional decision-making across time. And that's something no amount of prompt engineering has figured out how to replicate.
Tool spotlight
If the consistency argument resonates, it's worth looking at what the Lottie animation format offers for interactive work. A recent project combined Lottie animations with DOM events and distance-based maths to achieve precise, frame-level control over animated interactions in the browser — the kind of setup where engineering architecture serves the art direction rather than overriding it.
Lottie won't replace your creative judgement, but it gives you a framework where every frame is authored, not generated. For designers building interactive experiences who want that level of intentional control, it's worth exploring — especially as the temptation to hand motion design off to AI grows stronger.
Takeaway
Next time you're evaluating an AI tool for your workflow, don't just ask "can it make this?" Ask the harder question: "can it make this the same way, every time, on purpose?"
If the answer is no — and right now, for most tools, it is — then you've just identified exactly where your craft still matters most. Consistency isn't glamorous. But it might be the most valuable design skill you have.