Comment by AnotherGoodName

Comment by AnotherGoodName 13 hours ago

Fwiw this absolutely works amazingly well with modern coding assistants. “I want a diagram of equation X morphing into Y” or similar is always a one shot success for me.

Part of it is the simplistic syntax and the sheer amount of open source manim examples to train on but it’s a pretty great demonstration of ai coding agents time saving. Especially since the output video looking correct is all you care about here. Ie. I don’t actually care about the specifics of how my explanatory videos were created, just that they were created via a simple prompt and it gave me what i wanted.

sansseriff 10 hours ago

I remember listening to a podcast where Grant Sanderson basically said the opposite. He tried generating manim code with LLMs and found the results unimpressive. Probably just goes to show that competence in manim looks very different to us layman than it does to Grant haha

Reply View 5 replies

apetresc 10 hours ago

I wonder if that’s also because Grant uses his own branch of manim from which the mainstream public one (manim-ce) has diverged quite a bit.
I can imagine LLMs being very confused being asked to write “manim” when everyone talking about “manim” (and the vast majority of public manim code) is actually the subtly-but-substantially different “manim-ce”.

Reply View | 0 replies
AnotherGoodName 10 hours ago

I’m having 100% success even when doing transitions between screens etc on the latest agents. Wonder if this is due to time and the agents vastly improving lately. Possibly also grant knows manim so well he can beat the time to type a prompt. For the rest of us i’m tempted to make a website for educators to type a prompt to get a video out since it’s been that reliable for me.

Reply View | 2 replies
- sansseriff 10 hours ago
  
  If you're passionate about it, then go for it! Though be aware that several others have tried similar things:
  - https://www.befreed.ai/knowledge-visualizer
  - https://kodisc.com/
  - https://github.com/hesamsheikh/AnimAI-Trainer
  - https://tiger-ai-lab.github.io/TheoremExplainAgent/
  - https://tma.live/, HN discussion: https://news.ycombinator.com/item?id=42590290
  - https://generative-manim.vercel.app/
  No doubt the results can be impressive: https://x.com/zan2434/status/1898145292937314347
  Only reason I'm aware of all these attempts is because I'm betting the 'one-shot LLM animation' technique is not scalable long term. I'm trying to build an AI animation app that has a good human-in-the-loop experience. Though I'm building with bevy instead of manim
  
  Reply View | 1 reply
  
  icelancer 5 hours ago
  
  I have used Tiger AI Lab's harness and have several issues/tickets I opened there. It's not a very serious project and doesn't work... basically at all. It's a good idea especially with VLM review but the results are not good whatsoever.
  
  Reply View | 0 replies
icelancer 5 hours ago

Yeah I've mostly had Grant's experience. Some frameworks have hooked in VLMs to "review" the manim animations and drawings but it doesn't help much.

Reply View | 0 replies

pkoird 11 hours ago

Curious to see how it'll combine with a RAG on its documentation.

Reply View 0 replies