Chrome Brings Live-Captioning To Any Web Audio Source (arstechnica.com) 16
An anonymous reader quotes a report from Ars Technica: Google is officially bringing its "Live Caption" technology to any website with the new version of Chrome. The feature, which debuted on Pixel phones and should be available on most Android 10+ devices, lets you easily apply Google's speech-to-text technology to any audio source, making it simple to get closed-captioning on audio that's lacking in the accessibility department. Starting today, Google is beginning to roll out the feature to Chrome 89 and up on desktop PCs.
You can enable the feature from the Chrome settings by going to "Advanced" and "Accessibility" and then turning on "Live Caption." Live captions appear on webpages as a gray box that fills with text as the video or audio plays. You can drag the box around so it never gets in the way, and you can even pick between two sizes. Live Caption will attempt to work with every audio source on the web; you can temporarily close the box each time you load a page, but there's no way to enable it on some websites and disable it on others. Google says all the processing happens locally on your device and won't end up on the Internet. For now, Google says Live Caption "currently supports English and is available globally on the latest release of Chrome on Windows, Mac and Linux devices and will be coming soon to ChromeOS."
You can enable the feature from the Chrome settings by going to "Advanced" and "Accessibility" and then turning on "Live Caption." Live captions appear on webpages as a gray box that fills with text as the video or audio plays. You can drag the box around so it never gets in the way, and you can even pick between two sizes. Live Caption will attempt to work with every audio source on the web; you can temporarily close the box each time you load a page, but there's no way to enable it on some websites and disable it on others. Google says all the processing happens locally on your device and won't end up on the Internet. For now, Google says Live Caption "currently supports English and is available globally on the latest release of Chrome on Windows, Mac and Linux devices and will be coming soon to ChromeOS."
Which devices? (Score:2)
The feature, which debuted on Pixel phones and should be available on most Android 10+ devices
How does one know which devices it will work on, and which it won't? My $90 Android stereo head unit has the feature, although I haven't tried it yet. Yes, I am a cheap bastard, but it went into a used truck that only had AM-FM so it's a big upgrade anyway. Hmm, looks like it's a $100 stereo now, maybe they decided to package it adequately from here on out. Anyway my stereo is an ESSGOO (yep...) The listing doesn't say which CPU it has, IIRC it's some kind of Allwinner quad-core.
Re:Which devices? (Score:4, Interesting)
It's down to the manufacturer if they enable it on their phones. The Pixel 5 has an upper-mid range CPU so it apparently doesn't need too much horsepower to work. All the transcription is done locally, nothing sent to the cloud.
Re: (Score:2)
So is this something you can enable by fiddling a setting in build.prop? Or do you have to change your device fingerprint?
Re: (Score:2)
It might not be in ASOP, I don't know...
Re: Which devices? (Score:2)
I turned it off (Score:2)
I first noticed it when some of my streamed video had an extra caption box... Covering the subtitles. It did not do well with Japanese, and only OK with English.
I might look at it again when it can do translations.
Re: (Score:2)
If you've ever tried the captioning on a google meet call with some friends, you'll know it's laughably bad. Four British guys talking about the nothingness we've been up to in lockdown and it got maybe one word in 5 right.
I doubt this locally-hosted model will do any better, but good luck to them.
Neat but creepy (Score:1, Flamebait)
>"lets you easily apply Google's speech-to-text technology to any audio source, making it simple to get closed-captioning on audio"
While simultaneously agreeing that Google can spy on whatever it is you are consuming? (No, I don't use Chom* browsers, but I imagine there are some types of terms of service for that feature, and most will click "sure" to anything presented to them, if it even is presented. And if it is a live meeting that one person turned this on- will all the participants know Google is
Re:Neat but creepy (Score:4, Informative)
They can't use this to do that anymore than they already can. The captioning model runs on the device, it doesn't dial home.
Phones Home (Score:1)
You TRUST that a close compiled Chrome doesn't send back encrypted text transcripts from the browser? A piece of software that requires a lot of network freedom just to serve it's purpose that is so complex it's become enough of a platform they have an "OS" distribution based upon it?
How would you know it was off? or your microphone used as input? I could think of some great uses for parental child monitoring...
Re: (Score:2)
Sure. I don't trust Google very much, but I assume when they do their sinister business, it won't be in ways that can be disastrously exposed with a simple disassembler and/or packet sniffer.
Funny times ahead (Score:3)
Has anyone tested this with something like youporn or redtube?
Automated Slashdot Response (Score:2)
As a company, Google, like all the rest, is bad. In fact every human is bad and dumb and wrong!!11 not me though, I'm perfect and smart!
I hope they raise the quality bar (Score:2)
Having anything is probably better than nothing, but at least right now the auto-captioning's quality is annoyingly poor.
Re: (Score:2)
MS built okay local speech recognition into XP for free two decades ago - and that was version 5 of what they gave away. I wonder what the quality difference really is.