I've been building prototypes of new AI learning tools for months, but I recently learned that 3blue1brown open sourced his incredible math animation library, Manim, and that LLMs could generate code for it without any fine-tuning.<p>So I made a tool that automatically generates animated math/science explanations in the style of 3blue1brown using Manim from any text prompt.<p>Try it yourself at <a href="https://TMA.live" rel="nofollow">https://TMA.live</a> (no signup required)<p>or see the demo video here: <a href="https://x.com/i/status/1874948287759081608" rel="nofollow">https://x.com/i/status/1874948287759081608</a><p>The UX is pretty simple right now, you just write a text prompt and then start watching the video as it's generated. Once it's done generating you can download it.<p>I built this because I kept finding myself spending 30+ minutes in AI chats trying to understand very specific concepts that would have clicked instantly if there were a visual explanations on YouTube.<p>Technical Implementation:<p>- LLM + prompt to use Manim well, right now this uses Gemini with grounding to ensure some level of factuality, but it works equally well with Claude<p>- Manim for animation generation<p>- OpenAI TTS for the voiceovers<p>- Fly.io for hosting the web app<p>- Modal.com for fast serverless GPUs to render the videos<p>- HLS protocol for streaming the videos as they are rendered<p>Note: This is focused on STEM education and visualization, and it is particularly good for math, but get creative and try it with anything! I used it recently to teach my partner's parents a new board game in Mandarin (which I don't speak!)<p>I'll be around to answer questions. Happy learning!
As usual with these things: it is impressive that stuff like this can be generated so quickly at all, but the content is very superficial and often wrong or at least misleading. It's unusable for learning, but great for spamming platforms, just like NotebookLM for instance.<p>As an example, I asked about the Cantor function. It generated a 1:24 video, which is laughably short, explained correctly how the Cantor set is defined but showed a flawed visual representation, then simply skipped over how the Cantor function is constructed and simply states the basic properties. Sorry, but this is garbage content.
Those videos are apex quality videos. You might as well ask for Nobel prize literature quality essays from ChatGPT.<p>You can probably imitate the structure/scaffolding of a 3b1b video in a cargo cult way, but you are losing domain expert level verification of quality (which is why AI fails, because it's not a domain expert).<p>So heres how I'm hearing yoir question, and it answers itself: "how do I get domain expert quality from a non-domain expert AI?"....
Whose voice are you 'borrowing' for this?<p>Compared to <a href="https://www.youtube.com/watch?v=24GfgNtnjXc" rel="nofollow">https://www.youtube.com/watch?v=24GfgNtnjXc</a> this video is absurdly limited <a href="https://tma.live/video/9c8e725e-ec21-41a7-984a-317d84216497" rel="nofollow">https://tma.live/video/9c8e725e-ec21-41a7-984a-317d84216497</a>
Well, I'm blown away. "Show me how information propagates through a neural net."<p>I feel like this is the one thing that's been missing from all the LLMs: actual visual explainers, whether image or video. python plots only get you so far, and all the Diffusion stuff is nonsensical. This is amazing.
I have to give you props for not requiring me to sign up, I’ve seen many ShowHN posts lately that require me to unnecessarily create an account which always prompts me to close the tab immediately.
I tried a few times and this is my experience<p>1. Doesn't work at all on Firefox 133.0.3 (64-bit)<p>2. Works on Chrome 131.0.6778.205 (Official Build) (64-bit)<p>3. No existing links do anything but a sub second "Generating" which disappears quickly<p>4. Does not work in Incognito on Chrome 131.0.6778.205 (Official Build) (64-bit)<p>My prompt kind of worked but ended at 48 seconds<p>Prompt: "Describe a honeybee learning which flower to land on in a field using Markov Decision Process with Bellman updating":<p><a href="https://tma.live/video/88f535b5-0e5f-41ca-9bd8-e35e7aa8a95a" rel="nofollow">https://tma.live/video/88f535b5-0e5f-41ca-9bd8-e35e7aa8a95a</a><p>I ran it a second time and got a longer video of 1:55 but it primarily just created Text. It also didn't explain Bellman's equation and wrote it incorrectly:<p><a href="https://tma.live/video/88f535b5-0e5f-41ca-9bd8-e35e7aa8a95a" rel="nofollow">https://tma.live/video/88f535b5-0e5f-41ca-9bd8-e35e7aa8a95a</a><p>The second prompt kind of worked but ends at 47 seconds and then loops the final 4 seconds forever.<p>Prompt: "Describe how the OODA Loop, MDP and SPA learning approaches are the same"<p><a href="https://tma.live/video/ee7b5048-3fde-4f1a-8ec1-c8bb48883c75" rel="nofollow">https://tma.live/video/ee7b5048-3fde-4f1a-8ec1-c8bb48883c75</a><p>Overall this worked as described. It's more than fast enough, but fails to deliver on consistency and graphics.<p>A few more iterations and fine tuning and you'll have a solid Beta. I can see this being very useful after another year or so of use and tuning.<p>Great work and congrats on shipping.
MathGPT also has this (exactly in the same 3blue1brown style, so I guess they also use manim), and in my experience it does actually work better and tries to explain math and write the equations.<p>I think they use some extremely cheap model for writing the code, probably 4o-mini or similar.
Whether I click an existing example or type in my own it doesn't seem to work. A dialog pops up for a second saying 'generating video' and then disappears.
Most hilarious is <a href="https://tma.live/video/8eb2d318-3217-4c09-a8aa-3fc7e8bb7cca" rel="nofollow">https://tma.live/video/8eb2d318-3217-4c09-a8aa-3fc7e8bb7cca</a><p>I asked a history question - tell me about Reddy kings rule. It made up a physics rule and started talking about electrons.
The Manim is not great, words overlapping etc, the page itself needs a lot of work, generating and then nothing happens, and it seems in a lot of case your backend workflows have issues, with descriptions only starting and ending after a say 30s when it needs to go another 1-2 minutes at least
Pure Garbage: <a href="https://tma.live/video/5fe506ca-3831-4ba3-b9c6-ff899c571bf1" rel="nofollow">https://tma.live/video/5fe506ca-3831-4ba3-b9c6-ff899c571bf1</a><p>Although it is pretty impressive for what an LLM can generate these days.
Wow, this is awesome! Thanks for building. I didn't realize there was a protocol for streaming while rendering, though I noticed sumo.ai doing something similar for audio. Gemini with grounding is new to me also, very nice!
Haha that’s funny, the beehive one lays out the hexagons as if they were squares - so they overlap and have empty space lol! But still it’s a promising concept.<p>Btw for some reason on iOS I had to download to view the video