Project Naptha automatically applies state-of-the-art computer vision algorithms on every image you see while browsing the web. The result is a seamless and intuitive experience, where you can highlight as well as copy and paste and even edit google chrome pdf merge translate the text formerly trapped within an image. Type in your email below and sign up for updates on this project.
Words on the web exist in two forms: there’s the text of articles, emails, tweets, chats and blogs— which can be copied, searched, translated, edited and selected— and then there’s the text which is shackled to images, found in comics, document scans, photographs, posters, charts, diagrams, screenshots and memes. This entire webpage is a live demo. You can watch as moving your cursor over a block of words changes it into the little I-beam. You can drag over a few lines and watch as a semitransparent blue box highlights the text, helping you keep track of where you are and what you’re reading. C to copy the text, where you can paste it into a search bar, a Word document, an email or a chat window.
If you stare at these three animated gifs long and hard enough, you might not need to read anything. The comic decries websites which arbitrarily hinder users from absentmindedly selecting random blocks of text— but the irony is that xkcd should count himself among the long list of offenders because up until now, it simply wasn’t possible to select text inside a comic. Depending on how you look at it, this can be seen as a bug, or a feature. This might be improved in the future, because certain parts of the Naptha stack do lag behind the present state-of-the-art by a few years. It usually takes some special software to convert a scan into a PDF document that you can highlight and copy from, and this extra step means that a lot of the time, you aren’t dealing with a nicely formatted and processed PDF, but a raw scan distributed as a TIFF or JPEG. Usually, that just meant suffering through the document, or in the worst case, printing it out so that I could scribble with a pen along, while I read.
But with this extension, it’s possible to just select text from a picture, attached to an email, or linked from a class action lawsuit overview. It’s even possible for files you have locally on your computer. Simply drag the image file over to your browser window. Allow access to file URLs” checkbox. 30 degrees or so— sorry vertical text, I’ll figure you out later! But with these types of images, the actual text recognition becomes somewhat of a crapshoot.
Because it’s one of those little things that simultaneously applies to everything and nothing at once, but it really lacked that little spark of life. But vertically this time, all the text metrics and layout analyses were precomputed. And blocks of text become selectable — so that future users who stumble upon the same image will instantaneously load the cached version of the text. Project Naptha is actually continually watching cursor movements and extrapolating half a second into the future so that it can kick off the processing in advance, the cursor changes when hovering over different parts, try it out: Highlight the text over the cat’s face.