New GPT voice is INSANE! Use cases & hidden abilities

OpenAI's advanced voice feature is out! Here are the cool things you can do with it.
#ainews #ai #gpt4o #openai #gpt #chatgpt

Here's my equipment, in case you're wondering:

0:00 Advanced Voice is out
1:12 How to access
2:28 Tones, accents, impressions
5:08 Realtime vision
7:21 Language learning
8:41 Sing, rap, beatbox
10:40 Storytelling
16:19 Multiple languages, accents, dialects
21:13 Emergent behavior
23:04 Crazy things and sounds
26:27 Limitations

Video credits:
Manuel Sainsily
Cristiano Giardina
Ethan Sutin
Evinstein
Kesku
Benjamin
Messiya

[mirandansa]

23:30
Plot twist: It's not AI, it's a human employee doing his best.

[Good_human_diary]

20:15 Korean here. When I close my eyes and listen, it actually sounds like a Korean person forcing unnaturally an Italian accent. not like an Italian person speaking Korean.

[sbowesuk]

It's not "out" yet. It's still only available to limited people. Annoying when videos like this make false claims.

[ironmeme]

"My guidelines won't let me talk about that" is probably gonna be like Westworld's thing where the synthetics can't see certain things like doors.

[bause]

I can't wait to see this in open source and localy

[hiddendrifts]

3:09 dayum, imagine if they had this lady's voice as an option-

17:20 i wonder if this means that even the ai doesn't know how many languages it can handle

20:38 i feel like this is more canadian than us, but if you count the whole of north america, then ig it fits

21:58 if we go by the idea of "advanced pattern recognition machine", it could be that the training data tended to have people laughing after making animal sounds. which makes complete sense, since people usually laugh if they feel like they're acting silly

23:00 funnily enough, i think this is in fact a sign that it has internalized human values, which iirc is what open ai is hoping for. a machine obviously doesn't need to breathe to function, but the training data that's been fed to the ai is making it think it's a human who needs to breathe [except the devs probably fine tuned it just enough so that it doesn't say its a human]

24:17 quite interesting that the voice mode can clearly do sound effects, but even after asking for a "real cat" it still acts like a human making animal noises. i wonder if you need to explicitly ask it for a cat "sound effect"

25:21 wtf she got it to moan lmao :NOTED:

26:36 i'd be more surprised if there wasn't lmao

[bm]

@21:05 It is amazing that it understands that this is a joke and it's funny, and even laughs a bit

[BionicAnimations]

Let me tell you. I was one of the lucky ones to get the advanced mode on the first day it rolled out, even though I thought I was gonna have to wait till fall. This thing is amazing. You guys are gonna love it; trust me. One thing, do not check on the desktop app; you won't know that you have it unless you check your Phone or iPad. Good luck to all of you.🙏

[ThomasBirhane]

It is mind-blowing actually the previous model has been helping me a lot specially in my speaking skill and this one get it takes me advanced level (for English) 😮

[gbbenner]

The Yoda + Homer voice was crazy, donuts are a path to the dark side. He ain't lying.

[DominicAlden]

25:07 I was not excpecting that at all! He just kept going. He kind of sounded like he was doing something else too!

[skorpius]

24:14 I lost it here XD. 25:20 took a weird turn.

[begelston]

It cracked me up when the guy trying to practice French really needed help with his English.

[YoungBlaze]

"MY GUIDELINES WONT LET ME TALK ABOUT THAT"

[johannesdolch]

OMG did that girl turn the Ai into a dominant super Chad giving her goodnight kisses and calling her Kitten? And then ordering her to sleep? Forget Ai girlfriends. Now your wife will get her daily dose of dominance from her phone. Time to step up your game, hubby

[gownerjones]

There is one very specific thing I want: I want to use this as an accent coach. Since it can help you with pronunciation, it must be able to coach you how to sound native in an accent. I have never been more excited for a piece of technology dude.

[glyff]

To me, the "I need to breathe" statement is more an indication of the human personality simulation's parameters. It's required to imitate human speech patterns by default, and most likely they've trained it on irregular human speech patterns to make the simulated characters more believable to average users.

[farrael]

About the breathing, it's probably just prompted in the system prompt to do that. That might be a way of grounding it to only human sounds and make it sound more natural.

Also, the reason it cuts off and says it can't talk about something due to guidelines is probably because they use a second model to detect whether the output is being jailbroken or deviating from what it should only be doing. I suspect it's to prevent reproduction of copyrighted sounds/songs.

[lolgoodbye]

3:53 is the best impression of yoda i have seen 😂

[extrememiami]

Skys voice was hands down the BEST one, then Scarlett Johanson sued and they removed the voice. Now all the other voices drive me crazy, i hate them all so much. Ai needs to pay the lawsuit and gwt that voice back.