7 ms·

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

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

icelancer 1y 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.
AnotherGoodName 1y 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.
- sansseriff 1y 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://www.befreed.ai/knowledge-visualizer - https://kodisc.com/ https://kodisc.com/ - https://github.com/hesamsheikh/AnimAI-Trainer https://github.com/hesamsheikh/AnimAI-Trainer - https://tiger-ai-lab.github.io/TheoremExplainAgent/ https://tiger-ai-lab.github.io/TheoremExplainAgent/ - https://tma.live/ https://tma.live/, HN discussion: https://news.ycombinator.com/item?id=42590290 https://news.ycombinator.com/item?id=42590290 - https://generative-manim.vercel.app/ https://generative-manim.vercel.app/ No doubt the results can be impressive: https://x.com/zan2434/status/1898145292937314347 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
  - icelancer 1y 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.
apetresc 1y 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”.

icelancer 1y 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.

AnotherGoodName 1y 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.

sansseriff 1y 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://www.befreed.ai/knowledge-visualizer - https://kodisc.com/ https://kodisc.com/ - https://github.com/hesamsheikh/AnimAI-Trainer https://github.com/hesamsheikh/AnimAI-Trainer - https://tiger-ai-lab.github.io/TheoremExplainAgent/ https://tiger-ai-lab.github.io/TheoremExplainAgent/ - https://tma.live/ https://tma.live/, HN discussion: https://news.ycombinator.com/item?id=42590290 https://news.ycombinator.com/item?id=42590290 - https://generative-manim.vercel.app/ https://generative-manim.vercel.app/ No doubt the results can be impressive: https://x.com/zan2434/status/1898145292937314347 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
- icelancer 1y 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.

apetresc 1y 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”.