Hacker News new | past | comments | ask | show | jobs | submit login
Mathpix: Convert Images and PDFs to LaTeX (mathpix.com)
52 points by amai on May 28, 2022 | hide | past | favorite | 16 comments



I used to use them two years ago when they offered 200 snip for educational accounts. I then subscribed for personal plan for 5$ when I was writing my MS thesis. It was useful and make me write quicker ( Still needed manual edits for most of the time). I then cancel my subscription when they put a limit on personal paid plan. Then they raised the prices of the (now limited) plan two times in the last couple of months. They also moved the free plan to be more of a demo. starting this month this ia their plans

Free Subscriptions: 10 Snips / month 20 PDF pages / month

Educational Subscriptions: 20 Snips / month 50 PDF pages / month

Pro Subscription: 5,000 Snips 1,000 PDF pages per month $7.99 / month and $69.99 / year

I hope to see more competition in this field.


Mathpix is a great product that perfectly fills its niche and I’m very happy it exists.

I wish some open source alternative existed. Even much less accurate would be quite useful. I know latexify but it does only single symbols.

Maybe it will be implemented as an extra language in tesseract someday.


This is an open source alternative that works quite well https://github.com/lukas-blecher/LaTeX-OCR


Wow, thanks! I had no idea it existed. I’ve searched for something like it not that long ago.


I tried this out on a paper I like to read, but which only exists as a scan: http://users.cecs.anu.edu.au/~bdm/nauty/pgi.pdf

Mathpix did better than I expected, but still no-where near well enough for me to give the results to anyone without some fairly significant editing. It would certainly speed up creating a latex copy of the paper, but for example whenever something has a subscript and superscript, the superscript seems to be ignored.


yep, same hapened. You have to extract the pages of interest before submitting, they have no page UI.


Sorry, I changed by comment when I independantly discovered I had to extract individual pages :)


I am very satisfied by their service, for an avid scientific reader it is invaluable. Not only the output is good (it's mostly right, only slightly more verbose than necessary, for example a^2 will be returned as a^{2}), but also the UX is excellent.

The only perk is that they significantly downsized their free tier some time ago.

I think there are some papers with open source deep learning models for this task but I never got to try one of them, unfortunately.


Mathpix is great, but they have become much more restrictive.

I had an old version of this software which I think allowed for unlimited snips but now I am limited to 100 for the free version. Plus this is also a closed-source proprietary product. I will look into the open-source alternative linked in this thread.


This is a lifesaver for collaboration with people who only use M$ W*rd. I like to do all typing and formatting in org-mode. When I need to copy over a big table I’ve done in org (which exports nicely to LaTeX), I’ll use this snipping tool to convert it directly to a word table. I also like it for when I’m lazy, instead of copying formulae by hand into my notes, I’ll just use this.


I think their testimonials are all made up!

(I still had a good laugh, kudos!)


A real one here, the tech is beyond expectations, the decoding of equations is super, even of scanned sources. It only faults at complex environments such as matrices or diagrams (interpreted as cloud-URL to bitmap). It also OCRs surrounding text. Couples nice with Typora Mathjax for an integrated wysiwyg-markdown+latex experience.


I think they left out the testimonies by Genghis Khan and Adolf Hitler ... something to the effect of how they could have conquered the Holy Roman Empire and the remaining part of Europe if they had this software. I guess they understandably didn't want to be tied to those characters.


FWIW, pricing information was almost impossible to find on a phone: you had to scroll all the way down on the particular service page (eg. https://mathpix.com/ocr), so much so that I've missed it originally.

There are no links whatsoever to the pricing, so only going back and forcing myself to scroll through the entire page, did I manage to find it.


Kind of unfortunate that a few of the demo pictures have a binomial coefficient and the resulting latex code uses a long/complicated array to express it rather than \binom. I'm not sure I'd want to use this for anything I'll want to edit by hand in the future.


When you use this it's usually because you don't want to handcraft the latex to recreate an equation from another source. It's unfortunate that it doesn't create the most plausible form, because it would help when you want to edit the equation but it's definitely not a dealbreaker.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: