OpenAI Launches NEW GPT4-OMNI aka “HER” (Supercut)

GPT4o is OpenAI latest launch, here's a supercut of the entire livestream.
Join My Newsletter for Regular AI Updates 👇🏼
www.matthewberman.com
Need AI Consulting? 📈
forwardfuture.ai/
My Links 🔗
👉🏻 Subscribe: / @matthew_berman
👉🏻 Twitter: / matthewberman
👉🏻 Discord: / discord
👉🏻 Patreon: / matthewberman
👉🏻 Instagram: / matthewberman_ai
👉🏻 Threads: www.threads.net/@matthewberma...
Media/Sponsorship Inquiries ✅
bit.ly/44TC45V
Links:
• Introducing GPT-4o

[@jon4]

It's impressive - finally I won't need to talk to people.

[@MartinBlaha]

😂

[@alvaroproanoaviles5293]

Actually, yours is a top comment

[@SFgamer]

Let's not forget that human connection somewhere down the line is needed. But avoid the wrong people.

[@DihelsonMendonca]

This already existed. I talk to Chatgpt Voice on my phone at least for 1 year now. 😮😮😮

[@DihelsonMendonca]

This already existed. I talk to Chatgpt Voice on my phone at least for 1 year now. 😮😮😮

[@Md.MoynulAsikMoni]

The voice tone shifting from normal to robotic was just amazing 😅

[@DihelsonMendonca]

This already existed. I talk to Chatgpt Voice on my phone at least for 1 year now. 😮😮😮

[@mandrews817]

@@DihelsonMendonca the difference is that the current voice model cannot sing, is restricted to normal conversations, and doesn't understand emotion. This model, on the other hand, is far more emotionally inteligent, as it can pick up the subtleties in your voice.

[@DihelsonMendonca]

@@mandrews817 Let's see that in the real world. I'm waiting for the new model either in android or on a desktop. 🙏👍

[@thereal_JMT_]

RIP rabbit

[@prakhars962]

and ai pin

[@AINEET]

They really need to open source that garbage but they are too greedy to do so

[@Jayc5001]

F

[@Grahfx]

not only. They just killed 99% of AI apps again.

[@qwazy0158]

Sadly, Rabbit was always dead in the water from the beginning

[@goodtothinkwith]

GPT-Her

[@casperd2100]

No!? I hardly know her.

[@MA-channel1]

@@casperd2100 "Her" is a 2013 American science-fiction romantic drama film by Spike Jonze. The film follows Theodore Twombly (Joaquin Phoenix), a man who develops a relationship with Samantha (Scarlett Johansson), an artificially intelligent virtual assistant personified through a female voice. en.m.wikipedia.org/wiki/Her_(film)

[@pixelwizardai]

Some manic will argue about assuming gender😅.

[@Peterstarzynskitech]

Soon I feel like Chat GPT is gonna be our home assistants.

[@MrChinkman37]

It already is, if you have a bit of technical and research capabilities.

[@ThomasJDavis]

Yeah that's something I thought of as well. Helping managing finances, diet and grocery lists, healthcare, tutoring for kids. But also, having it be an A.I. for the household that's not connected to the internet for security of personal, financial and health information. And whenever it needs information from the internet, it will communicate with a separate agent to give it that information.

[@Peterstarzynskitech]

@@MrChinkman37 Yeah but you have to have those technical abilities. Not everybody does. So once it's natively built into the code, soon you're going to be able to turn on the lights and control your smart home with chat GPT without needing any technical knowledge.

[@JohnSmith762A11B]

Yeah there are rumors both of an OpenAI/Apple partnership and of Apple working on small home robots. Really, Apple should have started working on that the day GPT-4 was released but Tim Cook is no Steve Jobs.

[@DihelsonMendonca]

It should be since 1 year ago

[@bradstudio]

I can't wait to get this and tell Chat GPT to talk to me in the voice of Mr. T.

[@jeffwads]

Wait your turn, F00l!!

[@GarrettGalloway]

NOOOOO, you trimmed out the "your face looks like wood" mistake. Dude at 14minutes that asks about his appearance/emotions didn't have the camera tilted the right way. The first attempt said "Your face looks like wood" and then he tilted the camera to more appropriately face him and asked again. Why'd this have to be taken out????

[@Statsjk]

No. It never said like that. I actually said "It seems like I'm looking at a wooden surface".

[@GarrettGalloway]

@@Statsjk so his face...his emotions...look like wood, yeah?

[@carlosamado7606]

​@@GarrettGalloway that would need to be asked tbh. I would be curious to know if it would stand on the affirmation that it wasnt in fact a face. Saying a face looks like wood and stating it looks like is looking at a wooden surface is way diff. So further questions would have been appropriate to know.

[@dalaun]

​@@Statsjk the way I took that was the camera was facing the wrong way

[@GarrettGalloway]

@@dalaun yeah, zoom back to that clip and it looks like he had his regular forward facing cam on and quickly switched to selfie cam...but not fast enough.

[@prayhardco]

She said “after many many months” .. remember when innovation used to take many many years. Scary 😮

[@crisgath3512]

I wanted a supercut and knew just where to find it. Thanks Matt. Always onpoint!

[@WretchedRitual]

This is crazy it's literally 'Her', any idea when this is supposed to start rolling out?

[@didiervandendaele4036]

Immediately except not the new voices yet !

[@slateisa]

In their website they said is coming in the "coming weeks".

[@Alice_Fumo]

I'm using GPT-4o right now! However the voice features are apparently not going to be so soon.

[@iuripeniche6028]

@@Alice_Fumo next weeks, not months

[@Jeff-66]

Premium users are already starting to get 4o (text only), free will get it over the next few weeks. Voice like in the demo ... in a few weeks, in alpha, premium users only, at first.

[@AphanFX]

I cant believe how fast and responsive it is.

[@anta-zj3bw]

I have 4o. There's still voice latency

[@gordonthomson7533]

Try interrupting it. You won’t be able to. I’m same - give it a little while. The full voice functionality isn’t with us yet. I don’t know if it matters but I’m in UK, London.

[@theraybae]

Thank you for this supercut! I just got back to my hotel in Lima and this is perfect timing.

[@nehemiahjuan950]

Thanks for cutting this up for us man. No way I woulda listened to the whole presentation in real time

[@themostlegendaryman007]

Defiantly interesting though it kind of makes me wounder if they are trying to hold back and sandbag gpt-5 with this.

[@Mikoto_401]

Just a reupload?

[@lukegallivan-smith3399]

Yeah, I was also hoping he'd parrot everything with his up-talking!

[@notnotandrew]

One thing it could benefit from is knowing how long to wait before it starts speaking. If somebody pauses for too long in the middle of a sentence, or takes a deep breath as Mark did on stage, it senses a pause and starts outputting. The time at which it chooses to speak should be determined by semantics and context, not just length of silence. If the model functioned in a truly on-line real-time fashion and had output tokens for a pause, that would take care of the issue.

[@robwin0072]

Hello, Good Evening. In the demo, it was mentioned a couple of times - that there is a desktop App. I have Windows 10, is the GPT4o App available for Windows 10? How does one download and install the GPT4o App on Windows 10?

[@emanuelmma2]

Simply amazing. Can't wait to use it.

[@gunjitkumar]

One step closer to The true A.I. Girl friend

[@cyc00000]

And thankfully only limited to nagging and complaining 80 times in a 3 hour window.

[@qwazy0158]

​@@cyc00000I feel a day will come when we'll have to spend tokens for nagging mode... Thankfully, I won't be around for that lol

[@temp911Luke]

Living in an imaginery world with imaginery friend...sounds like a mental illness to me.

[@klaymoon1]

Wow, to enable the access like this to all - how much server capacity does OpenAI have?

[@somerandomguy685]

Microsoft's global capacity 😂

[@6AxisSage]

Theyve only released text an image in and text out so far, so if anything this new model will save on resources for the moment

[@bdanuw]

I gave it a try with the kids and it was really impressive !! Thank you for the shortened video ... Any idea if the samples mentioned by the presenter are available anywhere yet please?

[@temp911Luke]

Are you Plus subscriber ? Im not and cant use it for some reason.

[@bdanuw]

I am but could not find. Will check again today

[@Grassland-ix7mu]

With this much expression, the good old insult "You sound like a robot"will to be reversed to "You sound like a human" 💀

[@raulgarcia6191]

Were they using an ios/android app or the web version?

[@oneeetwooooothreeee]

imagine having human robot with all the motors, sensors...etc... and integrating AI into this, and make it controls everything by itself without having to "program" each step / operation manually... that would be awesome :)

[@Copa20777]

Thank you once again for the timely upload Matthew..the fact it is free and it also has an API ❤

[@middleman-theory]

Guys..It’s just a mix of gpt-4, CallAnnie with the emotive inflection of the Pi app (albeit with realtime data), and any vision-based llm. The technology has been around for a while. It’s not novelty, just convenience having it all in one app. People acting like we’re about to achieve AGI or something. OpenAI has farmed its ideas from open source for a while now. I’m waiting for them to show me something original that THEY came up with that shakes the industry. Even Google beat them to the punch with much of this technology.

[@star95]

The GPT-4o real-time video interaction method is currently unavailable. Real-time voice response is also not yet rapid, so a physical connection is still required for mobile devices on demo site

[@OleksandrKucherenko]

Unfortunately it does not support mixed mode: input by text, output by voice in my headphones (will be great for public places)

[@keithprice3369]

Awesome! My app updated in both desktop and android. The video mentioned desktop gpt voice app but I can't find that anywhere on the OpenAi site or online in general. He was on Mac, I'm on Windows. Anyone know if there's a Windows Gpt Voice app and if so where to DL it?

[@sulracing9710]

windows desktop Version comes months later than the MacOS dekstop version

[@panpdx8919]

So I am seeing him use his camera to show chat GPT information right through the lens. Not seeing this ability on Android is that only for iphone at this point?

[@sunlight8299]

It's rolling out over the next few weeks

[@MyrLin8]

Thanks Matt :) ... and I can soooo identify Jon4 :) + when do 'us dev types' get the code ;)

[@techpiller2558]

I know people compain about the vocal fry, but Scarlett Johansson's voice is just magical. It's mature but also youthful, it's sexy, it's comforting, it has personality, it has sass, it has strength, it's fun, it's vulnerable, it's just so human and woman in so many ways you just gotta love it. The Her movie was really good also.

[@yudrapawar]

I will use it for improving my English

[@davidbraun7356]

it's already available on playground for those who want to test it now. I tested it out but, but it was not impressive. Latency was high at 3.6s for a simple query, and it answered incorrectly. Seems most of the new features in the demo are for chatGPT tho

[@sulracing9710]

It's also already available on the normal website for Plus users, and I feel a much faster answer time than the normal GPT4

[@cesarsantos854]

I already got access to it but I can't test it now. The only thing I can say is that the context window still sucks.

[@thomasproshowski7538]

Amazing, technology Miracle. My congratulations. 👏👏👏👏👏👏👏👏👏👏👏

[@senthilramalingam9500]

Looks like we can make GPT 4 O as virtual wife to have a two way communication...that's remarkable...

[@RoyalOwl12]

This is kinda scary and beatiful

[@kanwarsingh5844]

It's in early stages but it will get better in upcoming years and act like a normal human

[@TriggerCurtis]

Where is the Desktop App?

[@MONEYMAN5811]

I have a feeling it will not be nearly as helpful in real world use, but it’s exciting nonetheless. One step closer to open interpreter from OpenAI

[@Maq_Vega]

I wish silicon valley was still going, this would of made an amazing episode with different characters showing off the new features

[@MA-channel1]

Thanks, Matthew!

[@dennisestenson7820]

I definitely will not be talking to chatgpt if I have any choice in the matter.

[@Someone7R7]

This is incredible

[@cristian15154]

It's great, but is also bullshit, how come the new ChatGPT desktop app is only for macOS and not windows...

[@temp911Luke]

Do they say its free with limited cap ? Can anyone use it ? I cant.

[@freddiechipres]

Its available now on the website. Just got access.

[@DevPythonUnity]

the model yes, but there is no documentation how to input Audio and resive audio via api

[@temp911Luke]

Are you plus subsciber ? Im not and cant use it.

[@DevPythonUnity]

@@temp911Luke we,, i can tell that the model is smart but, they wont allow audio creation for now

[@banana_Udio]

when is it available?

[@sunlight8299]

It's rolling out over the next few weeks. Keep checking

[@sulracing9710]

On the website already check it out

[@banana_Udio]

@@sulracing9710 nope. still the old version

[@BThunder30]

Thanks for getting this out so quickly today. Open source local needs to catch up quick. I don't like that GPT is still so far ahead.

[@MentalParadox]

This is crazy. Some thoughts... I foresee a future where teachers and classrooms are a thing of the past and children are taught purely by their personal AI. Or at least they will help with the homework and studying for exams. Children will grow up to be even more antisocial than Gen Z and Gen Alpha, who only grew up with the internet and phones as substitutes for social interaction. Gen Beta will spend a lot of time "socializing" with their AI assistants. Expect the "AI partner" market to become absolutely HUGE and birth numbers to plummet even further below replacement. These assistants will also connect to our homes and cars ("AI, open the blinds please.") and pretty much manage our lives. Like Alexa but... actually good. On another note, a curious feeling I had during the presentation was that whenever one of the presenters would cut off GPT, it would feel rude to me. I'd feel like "let her finish!". Subconsciously, something inside of me is seeing it as human, and it's 100% the natural speech

[@jeffg4686]

And they STILL haven't announced socialism...

[@gordonthomson7533]

It’s gonna be necessary. Simply needs to be something like that.

[@thedudeabides2531]

@@gordonthomson7533Actually, I think it would be more ideal to keep the current Capitalism system and just have the government pay a certain wage to all the unemployed. That way, when ai takes over all the jobs, you won’t end up with half the country going homeless.

[@gordonthomson7533]

⁠@@thedudeabides2531 I don’t think it’s “the current capitalism” at all, though. Socialism of the 20th century didn’t fail (look at Norway, etc). I think your fear is Communism. That failed initially because of the human admin and bureaucracy required. After that it failed because of US sanctions. We’re certainly approaching a period in the near-coming years where the admin issue can be totally resolved by AI - the technological advances if we actually group together and stop getting hung up on greed will blow everything to date out the water… There’s crazy wastage in the existing capitalist model and it’s entirely incompatible with a world full of extremely effective AI agents. Last thing if have expected me to say a few years ago but I don’t see any other viable way. The alternative is miserable.

[@anta-zj3bw]

Latency is still there, guys

[@markmuller7962]

Guys check the OpenAI 10 or so short videos, this stuff is game changer

[@T___Brown]

This video was less than than impressive.

[@markmuller7962]

@@T___Brown 🤣

[@troyhetherington6744]

Umm… so why Am I still paying if now the latest and greatest is available for free? (Maybe they explain later in the video, haven’t finished it yet….)

[@spencerfunk6697]

When the big companies start eating off the open source companies that have been eating off the big companies for the last year

[@HUEHUEUHEPony]

name one

[@spencerfunk6697]

@@HUEHUEUHEPony open interpreter

[@spencerfunk6697]

@@HUEHUEUHEPony eleven labs

[@kiiikoooPT]

wait, this is the video from open ai, I want to see your feedback not the video again :s

[@user-kb3bf8bf9g]

imagine having the voice of red fox hello dummy for kicks and giggles😂

[@RonLWilson]

Wow!

[@stephanebaribeau7465]

Wow it is very impressive 😮😮

[@Nekomesha]

"teeeheee just talk to me like the CIA isn't listening, silly goose!"

[@jwb-221]

Great video! Plus, any chance do you know if that beautiful brunette works for OAI? And/ or her name? Wow!

[@paolomatteoomero8303]

Where is Ilya Sutskever? #freeILYA - let him speak

[@Jayc5001]

Here was my take from the OpenAI broadcast: “Bro is acting like this is new tech. We have had Neurosama for a whole year. The AI content creator with live voice and vision. She can even read, chat and sing. Old news tbh. But it does seem like it'll be really useful for disabled people. And it's also nice that this technology is getting wider adoption. The real big news is that we're possibly getting GPT-4 Turbo for free! That means GPT-5 that's probably relatively close. Overall today's release was a little bit disappointing. Small model, not really that powerful. But it's interesting in its use case, so the normies would probably love this a lot. It might lead to broader interest in the technology.” But I have already found some problems with my thinking and some questions that I have. Comparing GPT-4o to Neuro, was a shaky choice. My assumption that it's a smaller model than GPT-4 Turbo is based on the inference speed. That makes good sense except it could be because of architecture changes or fast MoE design. The second problem is me thinking that all of the information has to be stored on the model so it has to be large. Who said that it's just one model? It could use RAG and have internet browsing capabilities. Allowing for a smaller model but it could be trained for search and reasoning. The reason for me thinking that it could be multiple models is thinking about the models output. Does it always output both text and audio? How would output even work for a LMM? I have the feeling that this is some new architecture. They said it is end to end but that would mean the model not only understands audio but can output audio as well. How does voice work under the hood? That's definitely new from OpenAI. Since it's done on a phone and desktop someone should use a network spy to look for the outgoing and incoming voice files. You could probably learn a lot by analyzing them. One big update I see some people missing is the new tokenizer. They have some weird language tokenizer things going on that we will probably learn more about with the new API access.

[@Trackman2007]

Honestly it feels like they try to present Iphone 1, when there is Iphone 15++. It is so MVP

[@ernstmayer3868]

Nice but I won't use it. It's a privacy nightmare. We need such tech self hosted.

[@SuperFinGuy]

Exactly, we need an offline version and the ability to personalize it.

[@giupidelloglio]

I’m Italian and I can tell you it can speak better robotically than Italian, its pronounce was terrible but still great results. Thanks Matthew

[@suraj_bini]

Microsoft and chatGPT , I think they're going for a divorce. Co-pilot already exists on the desktop.

[@StuartJ]

Shots fired.

[@didiervandendaele4036]

GPT-4o I love you ❤ Would you like to marry me ? 😂😊

[@NakedSageAstrology]

I do not like that they are rolling out the new features for the free users first, it just seems like a big middle finger to the paying customers who have been waiting for premium features.

[@joe_limon]

I am a paying user and I got access at the same time.

[@rozburg]

Im a free user, and I don't see shit. hasn't rolled out to me.

[@francisco444]

Dude chill. Paid users will definitely get something extra. Free users should be prioritized because of accessibility.

[@TheLordSocke]

I do think they will give us a new model soon

[@ramze111]

If its free of charge, that means that "free" users are paying other way... data?

[@rogerc7960]

The audio features are for paying customers.

[@johndcyc]

No, they said they are for all.

[@CosmicCells]

@@johndcyc Nah, GPT 4o is for all, not the voice app

[@johndcyc]

I visited their website, and from what I read I think everyone gets voice. We shall see. I could be reading it wrong.

[@sethorlando]

wow 2 years ago this was just science fiction

[@nickfixit]

This is game changing.

[@kristianlavigne8270]

This has the potential to massively disrupt society, job markets, the economy…

[@WordsInVain]

Not really.

[@HarrisonBorbarrison]

The Siri Pro we all need.

[@heathermorelock8764]

Really trying to homestyle americana...interesting ❤

[@mrdevolver7999]

Still not available here...

[@sulracing9710]

in germany we have acces

[@mattpen7966]

i want smarter models before this tbh.

[@laptopworld7543]

Why does chat gpt give cover to genocide and do you intend on aiding further tyranny cover with even more models in the future?

[@abdullahazeem113]

Free access fantastic

[@edwinkania5286]

And I have been calling ChatGPT CHAD.....

[@SmirkInvestigator]

I miss Sky

[@BRATELLO73]

closed AI "bottom"!

[@Goggleboxing]

Thanks for retransmitting this, however the supercut actually diminishes the presentation by removing some of the emotive and informative (behavioural, non-verbal) content. I don't understand this obsession with removing people's natural behaviours from the presentation experience. None of us is so time poor that we can't exhibit some patience and respectfully watch the full character of a creative team including their individual mannerisms. You cannot do this in real life, it shows an intolerance that simply magnifies the hostility bred by the socially engineered Internet generation.

[@Kisuke686]

great, use it and literal sell your life. Local processing or local models to go!

[@666Counterforce]

What a beautiful little angel - so cute ❤ and she speaks italian ❤

[@JamesMartin2014]

macos? no windows? wtf

[@rmat007]

I mcan cancel my subscription and use the free versioon! Thankyou

[@ShadArfMohammed]

As a free user I can't see it on the model list, maybe it was a scam to grab attention.

[@hoppy1]

🤢Just remember kids, each time you use ClosedAI you're feeding a monster. If you're going to chat with a bunch of data in a novel way the data should be public and the conversation private.

[@vladi1475S]

Both chat GPT and the girl need to work more on their Italian pronunciation :P

[@jonathanjones2435]

im trying to imagine children growing up in a world... where virtual game characters speak like real people. gotta be careful. children are going to get REALLY attached to video game characters and worlds.

[@osingularista]

GPT5 será o Jarvis da vida real

[@Batmancontingencyplans]

Kinda came here for Matt but met with clone of presentation😢

[@naninano8813]

yeah so no reason to pay for your gpt4 anymore. you only get extended token quota for your 30 bucks

[@KasperSOlesen]

If that is the case you can just cancel the subscription. That seems like a great deal to me. Its not like we are losing anything because more is made free. The subscription just promises we get new stuff earlier, and the free GPT4 does seem to be available to paid users first, then later it will be available to free users. I do not know for sure, I have not tried the free version recently.

[@fourlokouva]

You still need to pay for the API, which is the only reason to use ChatGPT at all over open source models

[@PrincessBeeRelink]

I am impressed to say the least...gg

[@eevvxx80]

Ofc it will be so laggy 😂

[@dennisestenson7820]

And I thought pleasantries couldn't be more useless.

[@mallow610]

Wait so you just posted their content as your own?