Your lectures, made accessible.
It doesn't just transcribe — it understands the lecture. Equations spoken but not written, written but not spoken, it gets them both. Play the video.
hi so today i just wanted to showcase the derivation of something
i'm not going to narrate the math i'll just showcase the steps one by one on my screen
and yeah let's go ahead
this is the equation we're looking at
step 1 we're going to isolate the x terms so that's what i do
and then we're going to divide through
the next step is to complete the square
and the step after that is factoring the left side
it factors nicely
and now we take the square root of both sides
and then we isolate the x
yeah hopefully you enjoyed watching this process
Hi, so today I just wanted to showcase the derivation of something.
I'm not going to narrate the math, I'll just showcase the steps one by one on my screen.
And yeah, let's go ahead.
So this is the equation we're looking at: .
And step one, we're going to isolate the terms. So that's what I do: .
And then we're going to divide through: .
The next step is to complete the square: .
And the step after that is factoring the left side: .
It factors nicely.
And now we take the square root of both sides: .
And then we isolate the : .
Yeah, hopefully you enjoyed watching this process.
Try it yourself
A full MIT physics lecture with dense equations on the chalkboard. Inline LaTeX renders fractions, integrals, and Greek symbols as the professor speaks.
Were those equations added manually?
No. Every equation was extracted automatically — no manual editing, no human transcriber, no post-processing. Minutes, not days.
The AI can hear and see simultaneously. When a professor says “this” while pointing at the board, it knows what “this” is. When they write silently, it reads the board. When they speak an equation without writing it, it renders the math from speech. ADA-compliant output at a fraction of the cost and turnaround of human captioning.
How we compare
| Human STEM captioning | Standard AI captions | Chalkboard | |
|---|---|---|---|
| Cost per lecture hour | $180–300 | $1–2 | ~$0.25 |
| Turnaround time | 3–10 business days | Minutes | Minutes |
| Renders equations | No — plain text | No — plain text | Yes — LaTeX via KaTeX |
| Screen reader math support | No | No | Yes — MathML output |
| Visual equation extraction | No — transcribes speech only | No — transcribes speech only | Yes — reads the board |
| Synced to video playback | SRT/VTT only | SRT/VTT only | Interactive viewer |
Human captioning costs based on published STEM captioning rates. Chalkboard cost estimated from Gemini API pricing at standard rates.
The numbers
Tested against MathSpeech (AAAI 2025) — in multiple cases our visual extraction was more accurate than the speech-based ground truth.
Built for accessibility compliance
Chalkboard closes the gap between compliance requirements and what current captioning tools can actually deliver for STEM content.
ADA Title II digital accessibility
Federal regulations require accessible digital content from public universities. STEM is the gap — no captioning vendor renders mathematical notation.
Screen reader compatible
KaTeX outputs MathML alongside visual rendering. Screen readers parse equations natively — no special plugins required.
Searchable, not a PDF
The transcript is searchable, selectable, and keyboard-navigable — not an image, not a scanned document, not a PDF.