General Marketing
We Tested the Claim That the New Claude Makes Disney Movies. Here’s What Actually Happened.
A viral post said the new Claude can turn any script into a Disney movie. So we gave one 10-second cartoon script to Claude Fable 5.2, Kling 4.0 and Seedance 2.5. One take each. Seedance was very good, Kling was a close second, and Fable didn't deliver.
Last night a client-sized question landed in my lap.
Anthropic had just announced Claude Fable 5.2, a new model "designed to tackle the most daunting technical challenges," with "improvements in areas like coding, reasoning, and vision."
Within hours, a viral post made a much bigger promise: "fable 5.2 can turn any script into Disney movie now.."
Is that true? Can a language model make the film you want?
We do AI content creation at CS & Co. So we didn't argue about it. We tested it.
The Test
We wrote one tiny cartoon script and locked it:
*A round little grey donkey wearing a big green leaf as a hat sneaks up on a single red chili pepper in a sunny forest clearing. He bites it. His eyes go wide, his face turns red, steam shoots out of his ears, and he blasts off into the sky like a rocket, leaving a puff of smoke. A tiny hamster watching from the grass shrugs.*
Ten seconds. Three makers:
- Claude Fable 5.2 (code), the new model from the viral post. It wrote code to make its film.
- Kling 4.0, the newest Kling video model.
- Seedance 2.5, the newest Seedance video model.
Kling and Seedance got the exact same prompt. Fable got the same script and one rule: no AI video or image generators. Build it yourself.
One take each. No rerolls. Then we scored each film out of 60 on six things: following the script, the characters, the motion, whether it looks like a real animated feature, cost and time.
Watch all three side by side, then each one full screen:
What Happened
Here's my verdict. Seedance was very good. Kling was a close second. Fable didn't deliver.
Seedance 2.5 won: 46 out of 60. The best fur, light and lens work of the three. The red face and steam were the strongest moments, and the hamster's shrug was the funniest. It cost 70 credits and came back in about 5 minutes. We ran it at 720p, because 1080p would have cost 160 credits, more than our cap for the test.
Kling 4.0 was a close second: 44 out of 60. Real feature-film lighting and grass, and the hamster sat in the grass watching the whole time, just like the script says. The donkey's face wobbled a bit from frame to frame. It cost 32.5 credits, less than half of Seedance, and also took about 5 minutes.
Claude Fable 5.2 didn't deliver: 37 out of 60. Fable has no video model inside it. So it wrote code. Its code built the donkey, the hamster, the chili and the forest out of simple 3D shapes, animated them, added its own sound effects and rendered the film. It took 21 minutes.
It hit the beats in the script, in order: the sneak, the bite, the red face, the steam, the blast-off, the hamster's shrug.
But it looks like a mobile-game cutscene, not a feature film. Trees made of spheres. Smoke made of stacked balls. No fur.
A brand doesn't ship a beat list. It ships what's on screen. On screen, Fable's film doesn't deliver what a brand needs.
So, Disney movie? No. Not from this run.
That's not a knock on Anthropic. Fable is built for code and reasoning. It isn't a video tool, and Anthropic's own announcement talks about coding, reasoning and vision, not film. The Disney claim came from a viral post.
The Part Worth Watching
Both video models rushed the setup. The sneak barely happened. The bite was fast and hard to see. They spent their frames on the spectacle: the steam and the rocket.
If the setup matters to your story, write it into the prompt and check for it in the edit.
The Honest Asterisk
This was one script, 10 seconds long, one take per model. That's a quick test, not a benchmark. More takes could change the order.
The Lesson for Brands: Test the Claim Before You Spend
The viral claim didn't hold.
Use video models to make video. On this test, that's Seedance first, with Kling close behind and cheaper.
Use language models like Claude to write and plan: the script, the shot list, the brief.
And test before you spend. A viral post costs nothing to share. Believing one can cost you a production budget. That's what an AI content studio is for. We run these tests so our clients don't burn budget finding out the hard way.
The same discipline applies to the creative that runs on connected TV: test it, then read the results in your own sales before you scale it.
The Final Take
Can the new Claude make a Disney movie from any script? Not in our test. Fable didn't deliver the film. Seedance did, and Kling came a close second.
The win isn't the loudest claim. It's the right tool on the right job, tested before the money moves.
That's how we build creative at CS & Co., and it's part of the CS Method™. Want to know where AI content fits your brand? See if we're a fit.
Cory Poccia
CEO, CS & Co. Marketing Studio™