[Old voice, new model and app] Asking ChatGPT 4o (Omni) to laugh with the new Mac App
Vložit
- čas přidán 13. 05. 2024
- The new ChatGPT Mac App. Trying to get it to laugh. Looks like not all of the new voice features for GPT 4o are released yet.
- Věda a technologie
Per OpenAI:
"We'll roll out a new version of Voice Mode with GPT-4o in alpha within ChatGPT Plus in the coming weeks."
AND
"We plan to launch support for GPT-4o's new audio and video capabilities to a small group of trusted partners in the API in the coming weeks."
To answer some questions:
1. I am using my personal paid ChatGPT Plus account in this video. I get an error if I use my business paid ChatGPT Plus account. I haven't tried a free account.
2. The Mac app in this video is not on the App Store for me yet. I got the .dmg from the OpenAI forums (persistent.oaistatic.com/sidekick/public/ChatGPT_Desktop_public_latest.dmg).
3. The Android App and now even the iPhone app have the same features for me (on both my accounts now). At first my iPhone had 4o, but not the headphone icon that triggers voice mode. Now all apps are behaving the same for me.
4. I realize now that the voice conversations have been in the mobile app for a while. See: czcams.com/video/SamGnUqaOfU/video.htmlsi=BpEpOq8DWIf0j2Lm
I still think this video is nice because it shows the improved performance when using 4o as well as the new Mac App.
The link to the DMG says Page Not Found now
EDIT: Found a link. Now it just says "I do not have access yet" once I sign in with my Plus login :/
@@pmarreck Dang. Yea, I happen to have two accounts and only one works.
Point being it will be better and the video reference is just the bad version?
This is the standard voice feature not the new voice feature of 4o. Even the smartphone app has this. U see it because it stops with a click and thinking then answering. The new version don’t stop and always listening
Good point. And this model said it can't talk so I think you're right.
Right you are. This is the new app with the new UI. I’m using the new 4o model, but it doesn’t yet have all the cool emotion / tonal stuff yet.
@@RandroidsDojo maybe she just had bad mood and didn't want to laugh
@@catdogcom bahahaha good theory!
@@catdogcom Real GPT-4o is hyper-attentive and never has any bad days
I think those personality features may take a few days
Even if it's the old voice, it has a lot of spontaneity already. It seems the update added the ability to be flirtatious, sing and increase the voice intensity.
The reason this old GPT voice (also been available on mobile for more than a year now) is the way it is, is because it is actually just text, being read by a text-to-speech voice. What makes the upcoming "4o voice" such a HUGE turn is that it actually HEARS and PRODUCES audio literally. It does not transcribe your voice to text (hence the slow response from old voice), nor does it need to necessarily read text like a normal text-to-speech bot. It ACTUALLY has a sense of sound. It can hear your laughs, your mood through tone and intonation, or even differentiate the different people talking with you or hear dog barks behind you. It actually has an experience of sound and it will get you when you tell it to laugh "this way" or talk in "this speed" or sing in "this tone". All real instead of converting whatever sounds it hears to text. So, yes. This is FAR from the upcoming NEW VOICE.
Right you are. I’ll make a video when I get access to the new stuff. Can’t wait!
@@RandroidsDojo do you know when this is coming? I have only seen the "in the coming weeks" but never a real day
@@MarcioGianotti idk, but I’m checking every day! Will post as soon as I get the new voice features.
It sounds exactly like “I am a real person” voice
I like how the voice is telling you it can't make sound.
Because it is just text, being read by a text-to-speech voice. The upcoming NEW 4o VOICE will be a literal bot that experiences and produces sound.
Unfortunately most of the multimodal features for GPT-4o haven't rolled out yet, including the new voice features. At the moment GPT-4o can accept image and text as an input and output text.
Soon though, audio input and output and image output will be enabled for paying users, which will be included in the new voice mode (so GPT-4o will actually be able to hear you talk and it will, itself, generate a voice in response).
Is it for paid ones? It seemed to say that all functions will be free in a limited form, and even now the old voice is available for free
This is the old voice, not 4o
Not the new voice features, but it is the 4o model with the new UI.
How do I get the Mac app? I've been looking all over
Anyway, this isn't the GPT-4o that was used in the OpenAI demo that was trained from the get-go on voice; this one was only trained on written language
Mac app link in the pinned comment.
It’s GPT4o, it just doesn’t have the new voice features enabled.
Unless I misunderstand how models are trained.
Where can you download the Mac app?
I'm getting some shivers.
maybe the prompt should be something like "pronounce 'hahahahaha like a human' "
How did you install it? From the App Store? 👀
I found a link to the dmg on the forums.
persistent.oaistatic.com/sidekick/public/ChatGPT_Desktop_public_latest.dmg
@@RandroidsDojo THANKS🙏🏻⭐️
Do you have it already in your phone too?
I have it on my Pixel, but not my iPhone. Also only have it on my personal OpenAI account, not my work one.
@@RandroidsDojo THANKS
can we use it now
this isnt gpt4o new voice model this is the old one you can tell by the latency of gpt
What app is this?
The new a ChatGPT Mac app. See the pinned comment for a link to download. Only done accounts can login though.
this is their newest kinda tech? glad to see it isnt getting too uncanny
Only part of the new stuff. Full voice improvements not out yet. The stuff they demoed is much closer to the movie Her.
That’s not even the updated voice, my dude! It’s the same voice model from GPT 3.5 I don’t think they will release it
You’re right. My mistake. See pinned comment.
You have to keep in mind right now you're only talking to the text generator it's not until they come out with the new feature later that it will actually have a conversation with you
I feel like it is like something in-between. Better than it was, but not with all the new features.
The new voice feature itself has not been released yet,if you watch the whole presentation of OpenAI,they said that they will release ALL the features over a few weeks....
Infact... paid users(who paid for chatgpt4) do have chatgpt 4O but without the voice update..
And currently free users dont even have that lol...they have absolutely no access to chatgpt 4O although soon we will too
Just look at the latency,it is not new voice update,actual one would be instant
Hope this clarifies stuff.. :)
Not exactly. I think they haven’t rolled out the tones, but this is a partial roll out.
On one account I have 4o, but I don’t have the voice option at all.
On the account in the video I have the voice option, it just doesn’t do everything yet.
@@RandroidsDojo I don't think this is the new voice modality in 4o, this appears to be the prior text -> voice
@@davefellowsright you guys are. Looking forward to the full release.
I am a free user and I have gpt 4o in a limited form
i want windows app, not mac os bleh why
This is just Text to Speech, we should be striking to openai for teasing a feature & not really giving it. 😑
You can put the blame on Scarlett Johansons diva ways. They pushed the release of the new voice features back from 2 weeks to 2 months.
Testing 4o voices when it's not actually available to you? Errr... The last 15 seconds were decent with the filler words though, that was amusing. Seems like a silly post though otherwise (and misleading clickbait?) :) but good luck with your channel. Cheers!!
Are you sure? I think the conversational voice feature is new. Before you had to send an audio recording, right?
Obviously doesn’t have the new tone and personality yet, but I thought this was the new app with a new headphone icon for activating the conversation mode.
Hmm, maybe it was there for a while. My bad, I mostly use the web app so I didn’t notice this before.
Anyways, thanks for watching. I’ll post again when I get access to the new mode.
the headphone icon was gradually rolled out, so not everyone had that option at the same time, and yeah I think mostly the headphone speaking interface is just on the app, not the web interface. Sorry, I didn't mean to sound snippy before in my first post, my apologies... The CTO in the OpenAI presentation said that the "emotive" voice feature will be rolled out over the next few weeks to certain advanced teams (like influencers, researchers, the Red Team, etc), but it might be a few months before the general public gets the new emotional voice, also gradually rolled out. Ah, I see you commented this exact info in another message below, yep, "trusted partners". Indeed, I can't wait to get access to those new features either, thanks! :)
This is not chatgpt 4o
It is. There’s some confusion about 4o verse the new voice features. But this is in fact the new model on the new app.