I can't read as fast as your demo GIF. Just infuriating.

it’s all code, if you want you can make a pr with adjustments to the demo

I bought a Dell Precision 7910 2x Xeon E5-2687W v3 (10 cores, 20 threads each) with 32GB RAM and 512GB SSD for $425 including shipping. I found that Windows 11 Pro will recognize only 20 of the virtual cores/threads. I don't feel a need to upgrade to more expensive Microsoft OSs at this time, so I just run Ubuntu natively on that box, which recognizes all of it. Assuming used DDR4 RAM returns to more reasonable prices at some point, I intend to load that box up to the 768GB max.

Right. Now try background color #767676 on the body element and see how much better it is.

I'm always happy to see more innovation in this area. It'd be great if you could make your model, weights, and training corpus public (preferably under a permissive license) on GitHub. It'd also be great if you could run some benchmarks against the other similar tools in this area (I'm thinking particularly of Mathpix, Equatio, and Microsoft's math OCR in OneNote, Word, and Azure APIs. If you make your test corpus and code available I could set up the benchmarks for you.

I agree that it would be nice if the model was open weights and could run locally.

I have digitized almost all of my college handwritten notes, I would love to transcribe them, check them for errors, and contribute that as training data, but only for open weights models.

I feel like an open weights and even open training data model is just about inevitable here, because a lot of people will feel the same way you do.

Please add a license file. Thanks!

Will do!

https://ai.meta.com/blog/nllb-200-high-quality-machine-trans... https://www.youtube.com/watch?v=AGgzRE3TlvU

This is a screen reader design flaw.

I am just getting started with my own cross-comparison, would appreciate your list of considered candidates if you have it handy.

Just this week I was working on something similar but specifically for users who have disabilities, so they can more easily report issues to site owners. I also combined general annotation capability so other users (of my browser extension) can read their comments. And also compatibility with Hypothesis (https://github.com/hypothesis, https://hypothes.is), also using the W3C Web Annotation spec. I hadn't thought of the drag-and-drop bug metaphor; I like it. I had also considered recording mouse and keystroke events up until the time that the bug is marked, and then bundling those events (sanitized) with the bug report for more precise repro steps, but of course that's a bigger ask for the opt-in.

You managed to make a web app in 2025 that has almost no HTML semantics and almost no keyboard accessibility (so very clearly and deeply not WCAG-conformant). Gosh, I haven't seen that in quite a while. Mind sharing which tools you used to make it?