22 Comments
User's avatar
Andrew Sniderman 🕷️'s avatar

Time to lose these substack shackles and go big on YouTube

Daniel Nest's avatar

Stay tuned for "Claude Design Animation"-generated slop channel by my evil twin, "Leinad Tsen."

Andrew Sniderman 🕷️'s avatar

I don’t know what that is but hahahahaha anyway

Daniel Nest's avatar

Hint: Daniel Nest in reverse. Wait, that's not even a hint anymore.

Andrew Sniderman 🕷️'s avatar

Werdna Namredins! Dam that was really hard to type

Daniel Nest's avatar

Sounds like an ancient curse or a rejected Harry Potter spell!

Andrew Sniderman 🕷️'s avatar

Ok I had to make an app; crazy how good one shot prompts are now

Jenny Ouyang's avatar

Love it!!! You picked the perfect use cases to demonstrate the power of Claude Design :)

I'm curious, how much time did it take you to finish each job?

Still using Hyperframes most of the time, now I wonder if I should switch to native Claude.

Daniel Nest's avatar

Glad you liked it!

It went surprisingly fast. I'd say all four demos (including the Computer Use article) took less than one hour in total, mostly waiting for Claude Design to process/reason and then asking for additional tweaks (in the case of the Marvel example, it came out like that on the first try so I didn't ask for changes at all).

Definitely give Claude Design a spin, though I can't vouch for it being fully relevant for your specific use case. What did you use Hyperframes for so far?

Jenny Ouyang's avatar

That was even better than I'd thought! Thanks for sharing the details.

I use Hyperframes for making shorts and youtube videos (just starting...), it seems to work. A 1-min beat may also take 30 min to 1 hour to satisfaction. Definitely hope to shorten that time :)

Daniel Nest's avatar

Ah, awesome! Would love to see some examples of actual finished products you've launched with Hyperframes (mine are just sample mockups, really) and also to hear how Claude Design compares with it in your experience!

Jenny Ouyang's avatar

Oh haha, now I’m feeling cringe about my claims…. I went to check my approved products on youtube, they’re not as good as your videos!!!

(actually the problem is my talking, not hyperframes :p)

This is already a higher bar for me: https://www.youtube.com/shorts/lOj2cH5oCPE

Daniel Nest's avatar

Hey cool, looks like a great format to try Claude Design animations, actually. Quick shorts and plenty of room for some nice visuals to go along with the concepts you're discussing! If you end up posting something, feel free to share. I'd be curious to take a look!

Jenny Ouyang's avatar

Will do!

TheAiBuildGuide's avatar

Jenny, is the 30 to 60 minutes mostly the wait while Hyperframes generates, or the back and forth after, watching a cut and asking for changes until it's right?

Daniel's demos were fast because generating is fast, and he still asked for tweaks on three of the four. If most of your time is going into the revising, that's the part switching tools might not actually fix.

Mohamed F. Ahmed's avatar

There's a pattern here worth naming: design tools built on top of coding-native models (Claude, Cursor's design mode) tend to handle motion better than dedicated design tools like Framer AI, because they're reasoning about state changes, not just visual output. I've seen the same gap in three separate client projects this year. Curious whether you noticed the animations breaking down on longer, multi-step interactions, since that's usually where the state-reasoning advantage disappears.

Daniel Nest's avatar

Not sure what “longer multi-step interactions” mean in this context, but I found that Claude Design holds the thread well even when I ask for multiple edits in a row. I can basically fix anything that's off by being specific about what I want changed.

TheAiBuildGuide's avatar

If chapter- and element-level edits work by not having to think through the whole animation again each time, that would explain why most edits still work across multiple tweaks.

It would also mean the real breakdown case isn't 'many edits,' it's one edit that spans multiple chapters or elements at once, where none of the three scoped options quite fits.