Word Lens App is Like a Magical Visual Babel Fish

Clarke is a god!

That is an awesome app..I could definitely find this worthy going on any vacation

That is really, really awesome.

wow that is pretty freaking cool! I think the way it works is by tracing the text in the picture and matching it with the different letters and then just using the same kind of translation technology as babel fish (for example) and converting it and placing it on the sign in place of what was there. Pretty awesome

I've remember Intel did some demo like this during their CES keynote.
Forgot which year it was.

We have speech recognition, we have online translations, and we have text to speach programs. Why has no one mixed all of them together? I've wanted this for a long time. With powerful smartphones and decent 3G/4G networks, this has all been made possible, so why haven't we done it?

Good idea..

"We have speech recognition..."

We definitely don't have universal speech recognition yet. It works well in limited contexts, like customer support phone services, and pretty badly in general.

That being said, it's getting better, and it's only a matter of time.

I wonder if that'll be available on my windows phone at some point? currently pretty much nothing is

MMM nice but the translation is not to acurate.

So its like Google Goggles but not free and quite a bit later to market?

Text to speech and reverse works reasonable well but machine translation is only acceptable when working with isolated words, like in the sign example shown in the article, or when the individuals communicating in two languages are writing about a conceptual domain that is limited and which both are very familiar (e.g. two geologists translating information on petroleum bearing formations). Even then the sender must be careful to use syntax that he/she has learned can be translated reasonable well by the software they are using. The syntax filtering process can take place unintentionally--without the user's conscious effort--if that person has used the software a lot. If those same geologists tried to translate the Koran from Arabic to English and the Bible from English to Arabic by machine, they would end up with mostly unintelligible text.

I cam't wait to see what the translations actually come out like:

"Nolonger 6xAA Batteries animal chaser, Please do not try to break wire gauze, throw no donut at door and drop no dog's feces on sidewalk!"

Question: why do the screenshots in the above show the scene from 2 different angles. Were the examples above faked or does the app also change the point of view. On the tow away pics the translated shot lets me read the building that was previously covered by the sign. The Sausage sign appears to be from a higher viewpoint in the translated pic as well.

Just wondering.

I'd still rather have the little fish I could put in my ear.

The image you look at with Word Lens IS faked... You really should check out the video to see how this works. The program tries to, in REALTIME, scrape off the english text (or whatever is on the sign), leave the background color behind, and plant in the translated text instead. The screenshots show slightly different POV from frame to frame because, duh, the phone wasn't ever showing both languages at the exact same moment and wasn't being held perfectly still. If the POV were 100% identical, you would actually have more cause for concern.

The sausage and lemonade sign does look weird though, the keystone effect detection/matching obviously is imperfect.