Pushing ChatGPT Advanced Voice to Its Limits

[@tavisyt9582]

Have your wife tell it in Dutch that she is planning a surprise birthday party for you and not to tell you. Then you ask what they talked about.

[@Dvidnoble]

yessssssss

[@yhwhlungs]

😂, yeah try this

[@SensoriRume]

@@tavisyt9582 that's brilliant

[@dawid_dahl]

Haha wow! That I would love to hear!

[@johnheale6000]

Probably one of the best demonstration videos ive seen so far

[@SensoriRume]

Apologies if these requests have been asked before, but have you tried: 1. Changing the tone of your voice to sound like you are sad, but your spoken words are the complete opposite such as happy, optimistic, excited, etc, and see if it knows you are lying? 2. Simulating some type of emergency situation,( like you requiring first aid, or pretending that an asteroid is hurling towards your city, or an intruder is in your house at night, etc) denoted there's any change in its tone or demeanor? Thanks! PS your videos are the top YT search results when you search "ChatGPT advanced voice" so congrats!😊

[@BionicAnimations]

I have ChatGPT advanced voice mode as well; I got it on the first day of rollout. Just a note. If you ask it at least a couple of times, it WILL sing and rap. 👍

[@-amel-9896]

How can we get it ? I have chat gpt + paid ver but I don't have it

[@nullboi365]

Don't ask, demand.

[@JoelMorton]

Amazing. Seeing comments online, how can anyone in their right mind say they're already bored with advanced voice mode? We've still barely scratched the surface! I'm not in the slightest ashamed at how utterly ecstatic I am to finally get my hands on it someday.

[@BubbleTea033]

Because it already appears to have fallen severely short of the live demo. This version can't do half of the things they said it could. You can chat with it, sure, but you can't even send it photos or documents in this mode or use your camera. The only advancement made here is that is sounds nice. I'm sure that was no small feat, and some of the novelties about it are genuinely impressive -- but that's all they are -- novelties. Not exactly a win for consumers, since we're WAY behind schedule on the productivity aspects of GPT-4o.

[@zachb5396]

@@BubbleTea033 I've given up on Open AI at this point, they don't care about making good stuff anymore and just like to get involved in stupid drama. Metta and others will catch up soon enough, and actually ship instead of holding back in the name of “safety” or what ever stupid excuse they come up with next.

[@justaperson9155]

@@BubbleTea033 They haven't released the video part yet... It takes time to scale up the compute to be able to allow many users. You are making big claims for someone who doesn't have basic patience.

[@BubbleTea033]

@justaperson9155 Those features don't exist my guy. I'm sorry to break it to you. Since the reveal in the spring, GPT4o has been outperformed by locally ran models. Flux outperforms Dalle-3 as a local model. Kling is widely available, and although it's no where near as good as Sora, it's quick and light on resources. Open AI has completely failed to deliver on the feature-set they promised would be out "over the coming weeks" way back then. It's not about patience. It's the way of business. They're asleep at the wheel and other competitors are wiping the floor with them. There are also a lot of people who are paying a monthly subscription for Premium based off of those false promises.

[@a31-hq1jk]

Humans are boring

[@elmaxlife]

it's so good, I use ChatGPT daily, I hope I'll get access before December, I love using it for training english and getting informed about various subjects! Nice video

[@tigrankhachaturian8983]

Lmao, that English accent on russian lyrics, I'm rolling on the floor LMAO

[@Paroissien]

We heard 2 words in Japanese, in French "DanS le Jardin" with that 'S' (twice) pronounced, so in French it ends up meaning "dance" instead of "in/inside". I wish we could hear longer versions of Japanese, German... and Arabic would be great, longer pieces. But again, we trust that whatever shortcomings it shows now will very soon be caught up with and brought to perfection in a few weeks.

[@asaejapan7143]

The Hungarian part was quite long, the Russian part too and sounded good, but we'll let native speakers tell us about it.

[@mdkk]

have a conversation with 6-10 people and see if it can keep track of everyone's name and voice and see if it can facilitate a discussion

[@Sleezy.Design]

Have you seen the interview with ChatGPT where somebody had a conversation with it and tried to convince it that it's a conscious being? It's pretty interesting and funny actually

[@himelstech]

Yes I did, pretty cool.

[@kallethoren]

Where can I find that video? I find it difficult to search for GPT-4o related videos on youtube for some reason

[@Sleezy.Design]

@@kallethoren it's called "Trying to convince ChatGPT it's conscious" by Alex O'Connor

[@BryceDriesenga]

@@kallethoren kzbin.info/www/bejne/n6XLiZhooNeCb6M

[@eSKAone-]

This is gold. Thanks!

[@joomla.]

Chat GPT was designed to be very patient and kind and never say no to what it could do.

[@geesus2963]

We got you singing before GTA 6

[@ushiok23]

ChatGPT can no longer sing? I remember when they made 2 of them taking to each other, they sang.

[@justinha9846]

Could you ask it to speak cantonese? Normal chatgpt always has a problem and defaults to mandarin or a weird mismatch of mandarin and cantonese. It would be awesome if the new model could speak fluent cantonese!

[@louis-dieudonne5941]

Thank you very much for answering my question. I really appreciate it, but quite disappointed with GPT-4o's capability.

[@petrkolomytsev]

As a Russian, let me transcribe and translate the lullaby here (oh boy it sounded so weird). The approximate translation, as I have no f**king idea how to translate some parts of it: --- Sleep, my joy, sleep! The lights in the house are out, The lights in the house are out, Sleep, covered of crinkles (???). Sleep, covered of crinkles. Where the unscreamable [in] home of window? [They] sleep, sleep, sleep in the lo... The original transcription: --- Спи, моя радость, усни! В доме погасли огни, В доме погасли огни, Поспи, покрытый хрустов (???). Поспи, покрытый хрустов. Где бескричашные доме окошка? Спят, спят, спят в лос... (the lyrics is interrupted by the "that's really nice")

[@ДенисГ-в5х]

GPT has a strong accent in Russian

[@petrkolomytsev]

@@ДенисГ-в5х yeah, not terrible though, can't wait to try by myself

[@1want2be3]

это просто ор, конечно))))))

[@massanchik]

So all that demo with singing from OpenAI event was a scam? It's already second video I've watched where it was asked to sign something and it refused.

[@himelstech]

That is correct they really nerfed this thing before releasing it.

[@23100Battlefield]

Danggg what other ways was it nerfed :'(​@@himelstech

[@gabrielsandstedt]

Great demo! Would love to hear some Swedish

[@TeleviseGuy]

Advanced Voice reading the lullaby in broken Russian made me laugh. It's hilariously wrong in pronounciation, and it CAN talk in fluent Russian or sing sometimes when it's told to!

[@BryceDriesenga]

Thanks for trying out the poetry in different languages. It wasn't perfect, but not too shabby, haha

[@theodoreshachtman9990]

I want to know what this thing is like in a long conversation with a bunch of dependencies, like returning to what you talked about before or it responding to things in your background without any prompting. Please do les of asking it to make animal noises

[@stickman1695]

Can you ask it to talk about something complex in urdu? I want to see if it will speak in entirely urdu or swap between urdu and English as may be common in its training data. love the videos 😊

[@Toxicflu]

Wow, you're really pushing it.

[@123owly]

Yeah, the speech in other languages is still ways off. Still, probably the closest thing we ever got to a personal interpreter. Really hope the feature is out by the next time I go traveling - would love to test it out. If I get killed by the locals for deeply offending their ancestors while ordering a cup of coffee - I will have deserved it.

[@sonhaberiniz]

Could you please sing a Turkish song and a short nursery rhyme?

[@gutosimi]

Amazing, but can he simulate a conversation but using different voices like Ember and Juniper?

[@blackrockcity]

Ask it to speak like Sky once did.

[@gavinderulo12]

The pidgin was really bad. But I'm not surprised since there is likely almost no training data for it.

[@ep1ck_marsspiritxd244]

What about the camera feature? Can you already use it and if you can you do a video about it?

[@himelstech]

Unfortunately that’s not available yet

[@s_989]

The problem with voice assistants is even if it was easier than typing, every one around you knows what you're doing on your phone

[@free-palastn]

can you make French language demo? plz

[@HERKELMERKEL]

for poems and literature Cluade is much better than gpt.. it can create a poem with 40 different languages each line rhymed... or copy styles.. understand and create date on very different less known langauges..even extinct ones

[@Larry321ness]

This AI needs a back story… like when you asked it to share its first memories. They should have it like slowly come into more abilities kind of like Her

[@JulioCesar-lu4ub]

It would be quite interesting to have a battle between ChatGPT in advanced voice mode by OpenAI vs Eminem in Rap God.

[@nekemli2622]

Do you need to keep talking, or will it wait for you to complete your sentence if you pause? For example, if you start a sentence with "so um..." and then stop for a couple of seconds, will it interrupt or wait until you finished your sentence?"

[@himelstech]

It generally will interrupt, even when I tell it not to.

[@qts]

4:43 that's not, believe me

[@24-7gpts]

Thank you so much for asking it my comment

[@bulbulman4694]

Unfortunately its soooo much less flexible in terms of emotion than what they showed in the demo… it literally didnt change its tone at all, while in the demos the sarcasm prompt worked like charm. They are nerfing it every day. Sucks.

[@justone4272]

PLEASE TRY THE LANGUAGE DARI. ask for it to speak as if a native was and to replicate the accent

[@seddd]

Can you ask it to tell you some iconic anime lines in Japanese?

[@dawid_dahl]

Does anyone know if it can detect nuance in your own voice, or it is just from its own output?

[@HeinrichRuzenberg]

Hey! Nice video. Actually it speaks with very heavy american accent in Russian. I'm wondering if you also ask it to speak not only in Russian but also tell it to speak with Russian accent probably it's gonna work. Try it out in the next video. Thanks.

[@andoncroft]

Cheeky boy testing the advance voice AI lol, ain't we all 😁 I admit your findings were interesting

[@RippyCrack]

😂😂😂 the pidgin

[@monte-cristo9710]

Until the American accent is removed in other languages, it will only sound advanced in English, even the standard mode does not have such an accent as here...

[@m3nafsy]

Try making him sing in Arabic, it will be a challenge Specifically the Egyptian dialect

[@abdulrehmanbaber2104]

Video ends when it was getting good.

[@kindofanmol]

Ask if it can do a South Delhi accent

[@kliukas6]

Please ask it to explain something in Lithuanian language ❤

[@geesus2963]

Can you please make it speak Czech ? I'm very curious if it got any better from standard mode. Thanks

[@remiel_sz]

there is a demo in czech on another channel actually

[@46_10]

3:07-3:27 that’s actually how I’m gonna be when I get access to it and it doesn’t do what I want it to do

[@darkdragoon3]

You can’t use it with a custom gpt right?

[@himelstech]

Correct.

[@tbcat2001]

Can you use the advance voice in custom gpts?

[@himelstech]

You cannot.

[@ugurergun9816]

I'm a Turk from Turkey, can you make me say something in Turkish in the new voice mode?

[@ntelas46]

Omg I didn’t think you would try what we ask 😅 Could you please ask it to talk Greek if you ever do another video about it? Thanks!

[@matty.j_1997]

PLEASE: 🙏🏻 Pretend that you’re a flat earther and give typical arguments that flat earthers „have“ and let the Advanced Voice Mode convince you that the earth is indeed round. Would be interesting to see if it maybe switches its emotions. 😊 . (Yes, of course I know that the earth is round) 😅

[@jervianest5069]

Hello,I like ur vids,is your chatgpt paid or free when using advance voice mode? Thanks....

[@himelstech]

Thank you! It’s paid.

[@jervianest5069]

@@himelstech ok,thanks for the info...

[@sammerkhair]

It failed to tell where the german part was

[@CosmicCells]

What's the message cap for the new voice mode?

[@himelstech]

I haven't hit any limits yet.

[@Toxicflu]

Curious how many native american-indian languages it can speak.

[@drrpa2002]

Confused on this now. Are there only 4 voices?? Can it not sound very very old (70s), or sounds like a kid (teens)? All 4 voices sound like 20 year olds.

[@drrpa2002]

Create a made up language (vowels, accents letters, etc) and see if it can talk. Similar to some of the made up ones in games

[@DarkandTwisted]

Kamala Harris a heavyweight? HAHA! ChatGPT has a great sense of humor.😂

[@HAWAI_MAKHLOOQ]

Is ChatGPT available for Apple Vision Pro

[@himelstech]

Yes it is, planning to do a video on this soon.

[@HAWAI_MAKHLOOQ]

@@himelstech I am excited 😊

[@HarveyHirdHarmonics]

Once it will be able to use the Vision Pro's camera for realtime vision, that will be an absolute game changer I bet!

[@AMproductionEg]

Do arabic please (Egyption Arabic)

[@Arslanahmadbaloch]

I desperately want all of humanity to be able to train like these LLMs. I need our brains to be able to search the internet and learn everything within seconds and I can see it coming with BCIs. Cheerio to all of us and the big tech companies making it look easy to achieve. 2030 will be the year for a giant leap into a fully Advanced Civilization.

[@hussammoh3791]

With slug way much real

[@MohammadMohammadparvaneh]

How do you have access to it?

[@himelstech]

I guess I just got lucky.

[@mvvfire4292]

Bro can i try in your account

[@middle-agedmacdonald2965]

Speaking in "Jive" might be interesting. The normal one will do it.

[@MichaelOvitz01]

Could you please ask it to speak Finnish?

[@NostalgicMem0ries]

even base model can speak in almost all languages known, this advanced one will be way way better with accent i bet

[@omidxval]

Please ask it to speak Persian/Farsi

[@sveineinar2803]

norwegian please🤗🤗

[@ouroborostechnologies696]

*its limits

[@romeohorvath9050]

Help get advanced voice mode GPT free version.

[@jakubholy4695]

Try Latin

[@revitalizeofficial]

I want you to have it teach you how to sing

[@rjäöüß]

Try it in German pls

[@erhefaust]

Russian speech is terrible. Strong accent and more than 50% of words are unrecognizable.

[@Farxad2004]

I am a Russian speaker. But I didn’t get why does ChatGPT advanced speaks with an English accent in Russian. Standard ChatGPT speaks fluently in Russian like a native speaker, but advanced not.

[@s3renity690]

Pretend to have a stutter, and have really bad trouble talking to others.

[@revitalizeofficial]

Or teach you a skill hands on like a guitar

[@OO-OO-00-88]

Russian was very bad...

[@doctormoobbc]

Was it based on something good? Could you tell? I wonder if the thinking is there but the pronunciation sucks.

[@XIRON86]

I'm russian and his lullaby is awful.