Voice translations preserve speaker's tone, pacing, pitch—with SynthID watermarks for security.
See full article...
See full article...
Google announces Gemini 3.5 Live Translate for instant voice-to-voice translation
- Thread starter JournalBot
- Start date
With an actual watermark, all I need is a light source and a few moments to evaluate.
Who can view (?) these SynthID watermarks and when would they do so?
Upvote
19
(22
/
-3)
Now we just need to load this onto a touch-activated broach like item we wear on our chest. Something shaped like say, a delta.
Upvote
73
(74
/
-1)
Digital watermarks aren't always obvious like a traditional watermark. It's a borrowed term that means more than the literal definition.
However: https://deepmind.google/models/synthid/
Upvote
20
(20
/
0)
Now we'll finally be able to understand the aliens when they say "take me to your leader", and we won't think they said "I will destroy all of you".
Upvote
-4
(2
/
-6)
There's a limited dev API, but for most people, you upload suspect content to Gemini and ask for a SynthID check.
Upvote
11
(11
/
0)
Uhura and company are FRANTICALLY paging through old
Klingon glossaries, manuals and dictionaries.
UHURA
(subtitled KLINGON)
We art delivering food... things
and...supplies to Rura Penthe...
over...
Pause...
KLINGON VOICE FILTERED
(subtitled KLINGON)
Don't catch any bugs!
Upvote
8
(11
/
-3)
Google tools can identify them. I think the idea is that when you upload AI generated content to YouTube, send it to a gmail account, or put it on a website that someone views with chrome, those tools can detect annotate it as AI generated. Or you can uplaod stuff to their tool to check yourself.
Or at least that's the idea.
Upvote
10
(10
/
0)
EmphyrioDonk
Wise, Aged Ars Veteran
And in your case, probably, you will take them that lying criminal in the White house...and they'll decide to destroy you all, after all.
Upvote
-10
(5
/
-15)
Translation is one of the best use cases for AI. I second concerns about privacy with work meetings being translated on Google’s servers, but opening doors across cultures between humans is a huge win. This kind of tech is a lot easier for me to support.
Upvote
45
(45
/
0)
Unrealistic to expect most people to learn more than 3 languages, though.
Upvote
20
(22
/
-2)
Regarding the automatically applied watermark, it's an interesting example of how the whole AI thing is literally infiltrating everything in extremely opaque and chaotic ways.
The generated content here is not trying to impersonate someone, it's not intended to create novel information, not usable to deceive... Nevertheless the soundfile is now wearing the yellow "AI generated" star, which might give people a wrong impression.
Everything we see/read/hear is getting a higher and higher likelyhood of being either a hybrid creation (with an unknown mixture of human and AI) or completely AI generated.
That really creeps me out. It makes me feel exploited in a way, because there's more effort and energy spent on CONSUMING the content, than on CREATING it. Which is the inverse of what it historically used to be. Am I the only one who feels like that kind of content shouldn't have a right to suck up part of the limited time I have left here among the living?
The generated content here is not trying to impersonate someone, it's not intended to create novel information, not usable to deceive... Nevertheless the soundfile is now wearing the yellow "AI generated" star, which might give people a wrong impression.
Everything we see/read/hear is getting a higher and higher likelyhood of being either a hybrid creation (with an unknown mixture of human and AI) or completely AI generated.
That really creeps me out. It makes me feel exploited in a way, because there's more effort and energy spent on CONSUMING the content, than on CREATING it. Which is the inverse of what it historically used to be. Am I the only one who feels like that kind of content shouldn't have a right to suck up part of the limited time I have left here among the living?
Upvote
-10
(0
/
-10)
Any company with a google enterprise contract is already having Gemini record and take notes on most of their meetings anyway.
The result of this is going to be me speaking in english, and a human-ish attempt of my voice in another language coming out the other side. There should absolutely be a built-in disclaimer/watermark on the resulting audio file indicating that MrDweezil never said anything in german and that both the words and their vocal inflection were AI generated.
Last edited:
Upvote
18
(18
/
0)
Real-time translations are amazing but with all things AI I get concerned over the long-tail effect of lazy humans.
e.g., If you marry into an extended family that speaks a different language, I hope you would still take the time to learn their language in their voice, and not just stick earbuds in your ears and listen to a robot.
On the other hand, if you like to casually travel and visit other countries, what a great way to eliminate barriers.
e.g., If you marry into an extended family that speaks a different language, I hope you would still take the time to learn their language in their voice, and not just stick earbuds in your ears and listen to a robot.
On the other hand, if you like to casually travel and visit other countries, what a great way to eliminate barriers.
Upvote
1
(5
/
-4)
I wonder how many people who speak in their own language in derogatory terms about other people around them, are going to get caught out by this?
Upvote
3
(4
/
-1)
SubWoofer2
Ars Tribunus Militum
Intonation is a supple thing to get right, as my Chinese friends smilingly told me when I greeted them with "I am an egg".
Upvote
9
(10
/
-1)
So it took 70 years, 1946 to 2015, for Dick Tracy communicators to become real. And 48 years, 1978 to 2026, for Douglas Adams Babel Fish to become a reality.
Upvote
14
(14
/
0)
Upvote
2
(4
/
-2)
I'm on the bandwagon of thinking that this is pretty great. I am very proud of the time and effort I've putting into learning a second, and now third, language. And I try to use and practice those skills as much at possible.
But while I love travel and relish some of the challenges with finding your way in a new place and meeting new people, language is such a huge barrier to many. My parents, for example, would benefit so greatly from this and it would take so much of the stress out of going to unknown places.
And maybe it might lower some "foreigners == bad" rhetoric in some places I know as well, simply because the tourists can't express themselves in the local language.
But while I love travel and relish some of the challenges with finding your way in a new place and meeting new people, language is such a huge barrier to many. My parents, for example, would benefit so greatly from this and it would take so much of the stress out of going to unknown places.
And maybe it might lower some "foreigners == bad" rhetoric in some places I know as well, simply because the tourists can't express themselves in the local language.
Upvote
6
(6
/
0)
I've loved Star Trek ever since watching TOS every sunday on my aunt's B&W TV as a kid.
Never thought I'd see the Universal Translator happen in real life.
Makes me wonder what else is coming before I'll pass.
Never thought I'd see the Universal Translator happen in real life.
Makes me wonder what else is coming before I'll pass.
Upvote
0
(0
/
0)
So.. will this model run on-device, or require an internet connection? Particularly for travel, the former is much more useful..
Upvote
4
(4
/
0)
It can't translate a language it's never heard before, so it's not a Universal Translator yet.
Upvote
8
(8
/
0)
I take your broader point about AI, but this is a translator. I don't really want it being creative.
Upvote
11
(11
/
0)
Was just talking to the wife about foreign movies last night. I happily watch them, but she can't read fast enough to deal with reading the subs and watching the movie.
Wonder if this might be some sort of solution that beats the crappy dubbing that's the usual alternative? Still, and likely will always have to be, a delay but it seems close.
Wonder if this might be some sort of solution that beats the crappy dubbing that's the usual alternative? Still, and likely will always have to be, a delay but it seems close.
Upvote
0
(1
/
-1)
Interesting phenomenon, reading speed.
I'm a very fast reader myself and while I have a happy IQ a scientist friend of mine has an IQ that's quite a bit higher.
He is however a much slower reader.
I have no idea what causes the difference but obviously it's not IQ-related.
Upvote
1
(2
/
-1)
I remember an article from 30 years ago on the difficulties of translating idioms. The English phrase "out of sight, out of mind" was translated into Spanish as "Invisible maniac."
Upvote
3
(3
/
0)
I know Google offers private enterprise options, where AI inputs are purportedly never used for training. But they’re not key onsite, so I’m not sure how secured these exchanges are. Google definitely hasn’t done much work trying to keep data private.
Upvote
0
(1
/
-1)
You really want me to Google that? I use DuckDuckGo thank you, with AI turned very much off.
Upvote
-1
(2
/
-3)
macosandlinux
Ars Praefectus
Well, let's see what Ive/Altman's device will do. Also these translation LLMs will be part (or are already in China) inside earbuds - spying on your (or everyone around you) every word.
Upvote
0
(0
/
0)
It was indeed a more general observation, not directly related to this application.
Upvote
0
(0
/
0)