OpenAI GPT-4o API Explained | Tests and Predictions

OpenAI GPT-4o API Explained | Tests and Predictions

👊 Become a member and get access to GitHub and Code:

🤖 Great AI Engineer Course:

🔥 Open GitHub Repos:

📧 Join the newsletter:

🌐 My website:

Explaining the OpenAI GPT-4o API. My predictions and some tests of what I think we can expect from GPT-4o API and the multimodal model.

00:00 OpenAI GPT-4o API Intro
03:23 OpenAI GPT-4o Explained
06:47 OpenAI GPT-4o Exploration

[elliotanderson]

It's going to be a game changer if OpenAI can actually deliver all the functions they demonstrated.

[Ginto_O]

I just realized that new openai model killed your voice assistant projects, just as they did last time with GPTs

[mwkoti]

What we need ASAP is an open source alternative to GPT-4o realtime speech-to-speech (as in demos). I'm pro open-source and I want full control of the application flow, preferably offline. Has anyone tried to use XTTS streaming capabilities succesfully, for example by extending AllAboutAI-examples?

[Ouyk]

4:59 It's not horribly wrong, but I'd combine the <USER VOICE> and Voice IN in one graphic in the GPT-4o Voice API Now section and Voice OUT under the LLM RESPONSE. GPT-4o is going to be a game changer for education.

[YunusDogan-yclx]

Hi Kris can you do it vision version also with camera ? Some things or help usecases

[elliotnyberg]

I wounder how they manage to handle interuptions during the voice output like in their demo for the api

[Ms.Robot.]

This was very educational. Your instructions were clear and concise. ❤🎉

[OliNorwell]

That "Performance Scores" table - nice! If that is all correct then that's pretty impressive.

Though I did a screenshot test myself and it mistook a 3 for an 8 so it might not be flawless.

[alirezasheikh]

I think you're right. Maybe minutes after OpenAI presentation was done, I posted on their developer forum if voice in / voice out will be available to developers soon. They said only to a small group of "trusted" partners. So yea, I'm not sure when we gonna get access to this. You gotta be in that special circle. 😅

[ionutownprint]

Does the streaming audio function help with the latency?

[squiddymute]

multimodal or multimodel ? does really anyone believe it’s a single model ?

[ziadnahdi]

if you could give it IQ tests visually ! that would be great since this is difficult for it like how many triangles or what is the next humber in a series..

[athemis]

Hi, great video! Iam courios how much do those apis cost. On their website I found text pricing in tokens and it is pretty cheap and understandable. However the image or “vision” function seems to be so expensive. I calculated it and with full hd on low settings it is going to cost about 5$ for a minute on 15fps. That’s crazy, not even mentioning it their TTS, that costs 15$ per 1M tokens, which is pretty hilarious

[RaysAiPixelClips]

I made a script just like yours and now with GPT-4o it kinda defeats the purpose...😅at least we can still use it with local models.

[learnwithyan]

My thoughts that we already have a good chance to be so so devs and to write perfect code 🎉😅

[GamezeveR]

GPT5 may finally be able to tell how many r's are in the word "strawberry", but 4o will suffice with its ability to write a bigint from scratch in C in just a couple minutes of telling it to try again

[rodrigov.]

god the number of men who will fall in love with that voice hahaha

[babaksard]

why does the personality of the voice need to highlight the worst type of female personality type ie valley girl who laughs at everyone and everything for no reason tho ?