Show HN: My Free OCR Website – want feedback | Hacker News Reader

Show HN: My Free OCR Website – want feedback

myfreeocr.com

Having just done a bunch of research into OCR providers and encountered your site in the wild, I'll give some actual potential customer feedback on the site.

The main feedback is the visual design of the site is sufficiently non-professional looking (mid 1990's looking?) as to discourage any deeper evaluation of the product. I didn't stick around long enough to assess the product because your marketing materials communicated to this potential customer that your site was not worth looking at in more depth.

Marketing is hard and as specialized a skill as programming. The site at present radiates it's either an ancient or hobby site rather than a service provider customers can depend on. The lack of a visible business model makes the site seem even sketchier. Someone is paying for those compute cycles, so is this site in operation with the goal of monetizing the uploaded data? (As the saying goes, if you can't tell who is paying for the product, you are the product).

There were enough questions raised by that initial quick look at the site to lead me to immediately move on to other competing OCR providers.

As is the case with most tech startups, building the tech is just one of many hard parts of actually doing the startup.

So you knew about my site before I posted it here? How - where? Thanks for the honest feedback - that is exactly what I am looking for. I know the design is not the best and we have plans to make it more modern. If I will implement an OCR API then I have to support good service and the uptime must be good. Currently I have ads from Google AdSense on my site so there is a business model - to get great traffic and revenue from Google and then when the API is ready I could have subscription plans. What other OCR providers do you like or look professional? Yes, you are right the tech is just one part - that is one of the reason I am looking for feedback here. Thanks again - great to get your valuable feedback.

> Currently I have ads from Google AdSense on my site so there is a business model - to get great traffic and revenue from Google and then when the API is ready I could have subscription plans.

Yea just skip the ads and go straight to a subscription. It looks like garbage click farm sites. Ads are horrible and tech people disproportionately hate them. So if your goal is to make me think "is there an API?" then you can't look like that.

Look at Stripe/Plaid/etc to get an idea of what an API first sort of site would look like.

The big headline of plaid is "The easiest way for people to connect their financial accounts to an app" and the subtext is "get API key. click here". Don't bury your value proposition behind ads, especially when the ads google will place on your site are (a) competition or (b) spammy "CLICK HERE convert pdf now" type ads.

Visit your site via a public VPN on a clean browser without any google affiliation logged in or cookie'd. See what kind of ads actually show up. Do you want those on your site?

I am currently using Adsense settings that allow Google to choose the placements. It seems it makes the sites too crowded with ads - I agree. We are planning on a new design and when that is done we will have specific spots for ads and not allow Google to choose how many or where to place them.

I didn't know this was a thing. Google chose that ad placement? Does it vary from viewer to viewer (eg. does your page look different than mine?).

How much is the per-view income from a site that does that? This explains why a lot of crappy sites look like that - handing over control to google to maximize ad revenue income causes google algos to optimize to some maxima of low-denominator quality. Who would intentionally place full-screen ads, after all!?

I built something similar in the past using TesseractOCR and Apache Tika and PyPDF2 / QPDF. The idea is sound. An API based OCR already exists in Apple / Microsoft / and Google so I am not sure this would be that useful. There would be no way for the user to trust that you are not taking the data you are OCR'ing and using it. If you can apply some type of one way encryption of the content and prove it via open source code (like Whisper Systems does for Signal) which seems like overkill and lots of effort for a free app.

> An API based OCR already exists in Apple / Microsoft / and Google

Where can I reach them? Thx.

https://centraluseuap.dev.cognitive.microsoft.com/docs/servi...

We have build a web client + REST API that allows to use the API for free for small personal projects.

https://konfuzio.com/en/ocr-api/

It supports handwriting, correction of HOCR text via the webrowser, automated language detection.

We use the text to allow large enterprises to train document categorization and data extraction AI in a low/now code UI.

Disclaimer: I'm one of the founders.

Thanks, I will check this out. Nice to explain the background also.

I own this website MyFreeOCR.com you can use it to convert images to text. It supports 100 languages and many different formats (pdf to text, tiff to text, bmp to text, jpeg to text, bmp to pdf, jpeg to pdf, and more. I would like to get feedback and feature request from you. Currently we are looking into developing an OCR API. Would that be useful for you?

> Currently we are looking into developing an OCR API. Would that be useful for you?

Oh yes! Absolutely! However, I only would use it for private purposes, I do not plan to write an application for it, that I'd release. I would do a lot of preprocessing (mostly cutting into boxes, which I would then save in separate bitmaps) and then upload that.

Just these days I am in the process of converting the scans of a recipe-book, that is not available electronically, but which I want to have on my smartphone. There is a PDF online, but that is only a slideshow of bitmap images. It also has ornamental frames on each page, ornaments in text, etc. And as I did not find the results of open source OCR apps to be satisfactory, I started roaming the commercial sector (I still have to evaluate ABBYY Finereader, but that would be very expensive).

Ideally, so I thought, a household would have an account, that is realized via a basic subscription plus pay by volume, and then available as a REST API. The subscription provider would offer a simple user client like https://www.roxyappsdev.com/applications/windows-10-applicat... and a mobile client (think "document scan", with batch and auto deskew/process) but otherwise work (hard ;-)) on the recognition. Now, if hand-writing would also be possible, that would be great!

BTW: What is 'OSD'? And why not some graphics, that explain the different page-selection/-recognition modes?

> Currently we are looking into developing an OCR API. Would that be useful for you?

The above question is different from the below question:

> I own this website MyFreeOCR.com... I would like to get feedback

The feedback I have is (1) website is garbage - too many ads. Clearly not targeting devs who might want an API. (2) OCR API seems like a cool product. Focus on that. Privacy of my data if I call API? How is data trained? What language/alphabet/etc? How do I try it out ? What formats?

Bonus feedback, the best dev API sites are easy to see value proposition and easy to find API key + docs. Count clicks required to get API key + find template cURL copy/paste. And count discrete pages. Even if "coming soon" - show me the critical path today so I can find it later. If you want to start with an "upload a pic" site, still do above, but structure the upload tool as a "try out the API without needed to leave browser" where you upload a photo and can tune other params via web form.

PS. Lots of OCR api's exist, so it could be a cool project but what is your goals here? Do you want to startup? Why your tool and not BigCloud?

I guess if you write on the website that you promise to delete the documents afterwards, its OK...

We delete the documents and state that on the website - we only count the chars that we convert and add that number to the total number of chars we have converted from the start of the website.

Yeah, that's as believable as Facebook saying they can't read your messages.

I am not sure how we can provide evidence that we don't store those docs unless it is just a desktop app that works without internet connection.

Our big problem is client confidentiality, so we couldn't put PDFs over the internet. An in-house standalone version (maybe you could charge for support and updates?) might be just the thing.

Can you elaborate on the options for doing that outside of a desktop app? Like a standard web app accessible only via company vpn?

That could work, I don't know what the backend software is, maybe node.js. Our in house web apps are built around Telerik but I don't know much about what it does.

A few things (as a user of Google's OCR API):

* How does this compare to other OCR tools on standard benchmarks? * I don't see a terms of service, you might want that. * The privacy policy also seems to be lacking. IANAL but you should hire one to make sure that stuff is squared away.

Thanks for the feedback. We have privacy links: https://www.myfreeocr.com/en/privacy - we have to look into the terms. I haven't done full comparison to other tools yet but we seems to support more languages and image/pdf formats than other tools.

I like it. Two small bugs.

First, it is slow. Slower than the Tesseract app on my old Android phone.

Secondly, the txt it produces is misformed. With â€™ appearing instead of an apostrophe.

I'd recommend displaying text text on the website rather than a downloadable txt file.

Bit, overall, pretty good!

Thanks for the feedback. I agree - it is on our todo list to look into the speed issue. Will check out this issue with the apostrophe as well. Good idea to display the text.

Reply on news.ycombinator.com