chrome only | Analyzes video frames for text using the Canvas element to read pixel data. We’re doing edge detection, to split the characters and a crude OCR based on ASCII signatures.
Chrome only. Seems like Safari might work as well.
Node Canvas, socket.io,
Voting is now closed.






















(23)
trycatcherWow!! OCR for video is huge!! Imagine that for ads!! Seeing great potential here! But I wish I could to my own video to see how good the algorithm is. :/ Maybe live video? Even with the video provided, we got scrambled words. Maybe this ideia could be kept to the next nko. There’s not much aestetichs, just one page but it not ugly. Easy to understand. | |||
very cool idea but doesn’t seem to be very complete. once with more features, i can see it being quite useful | |||
therazorbladeThis is a really cool idea but missing a lot of features to actually make it useful. Even in the examples, it is completely focused on the bottom scroll and ignores all the other text displayed. The ability to tell the parser what to focus on, take snapshots, and actually make the translation work are really important. | |||
This is a fun entry with cool potential, but… I’m missing the part where this is a Node.js-centric entry. I feel I have to dock points on this; the competition was to exemplify Node.js and features related to it, not to build a (rather awesome) <canvas> utility. | |||
The design needs some work to indicate what is going on and how, but overall this is quite an awesome project and could have many uses. | |||
Excellent idea, but the translation needs a lot of work. | |||
mattsoldoThis seems really cool – although it would be helpful to have a little more info on what it is supposed to be doing. I’m assuming its OCRing the closed captioning? If so, great idea. | |||
very nice, i’d love to see this extended to support custom video URLs, and different colors and fonts. i could see this as an app for detecting ads and muting them, or as an automatic subtitle extractor. | |||
patmeenanThe word and character finding look to be working fine but the translation isn’t working at all (even for letters that look like they should be easily detected). Has promise but it also looks like it’s being hand-fed the best case scenario with the words in the streaming ticker – would be more impressive to do word finding in the full frame. | |||
waxpancakeGreat demo. The text recognition and letter splitting works well, though your character recognition was almost never able to properly guess the letters. Still, very innovative prototype. Keep it going! This would be huge with YouTube integration. | |||
fredyatesivCool idea. Not sure what the end goal is and the translated column confuses me but it’s neat to see technologies like this being able to make their way into main stream development. | |||
wow! this is pretty awesome, but how much is done with node vs the browser?