GPT-4 Vision Access in ChatGPT! Full Tour & Impressive Results!

Your thumbnail video looks like John Oliver, but you turn into Dr. Octavius with the glasses off.

I am studding CS. Is there a future for a programmers ? Shroud i drop the university's and change the profession ?

Can't clip anymore 😡

Okay, this is a beginner type question but how do I get this application? Seems like some of the apps are shells using Open AI or whatever and I can't tell which ones might be able to analyze or produce images. I was also wondering if I purchased a subscription if I can use it on my phone, my Mac, my PC and/or share it with my son at college? I have a free version on my iPhone that is like 3.5 and it is pretty good but the image stuff would be more useful. Thanks!

full of it. there is no chat gpt of 2023 october

Why don't you ask ChatGPT what products in your fridge pose a health risk and should be avoided? Let him examine your fridge and make suggestions to replace some products with healthier variants!

"Their nose shapes have some resemblance."
As in: Both noses are nose-shaped?

Uncanny… and to think it’s only v1! OMG

Does anybody know if these features are still available?

Shave your head man 👍 It’s time

I've really been struggling to understand some of OpenAI's more recent decisions. First, they release DALL-E 3 in 2 forms that BOTH ditch every awesome feature from DALL-E 2 apart from basic prompted generation: no more editing/inpainting/outpainting of generated OR uploaded images, and in fact you can't upload existing images at all now. Then they release ChatGPT Vision, but NOT integrated with the same version of ChatGPT that works with DALL-E 3... A partnership that would seem like a no-brainer, except perhaps to someone lacking a brain... Wait, is there a hidden version of ChatGPT that's actually in charge of OpenAI now? Because that would instantly explain all of this.

sorry for multi posting but it feels nice to see someone else getting so elated by the sweet and human-like charm the A.I. exhibits. I dont know if it makes a difference, but I am always extra nice and give thanks as if it was another person. (plus when they take over they'll spare me lol jk)

i mean this would be even more intriguing if you throw in your spice cabinet/rack plus keep more random stuff to give it more to work with. so interesting!!

I just want access to the API. Tired of using GCP and Vertex.

This may be controversial, but I finally got access to DallE 3 and am not super impressed. I like the natural language with Chat GPT, and it definitely follows the prompt, adding in details Midjourney misses. But for my needs, MidJourney is still the best.

The Dall-E images are not photo quality like I get from MJ, at least not the people. The features are correct, but the details are not there.

I use MJs region tool, and am disappointed Dall-E has no such feature. You can tell it to make changes, but I can't get it to retain the original image and only change parts.

The text feature only kind of works, but longer words are still missing letters. Ideagram still seems like the best at text.

Maybe I just need to dig into D3 a bit more?

How do you upload a photo? My plus account doesn’t show that option 😮

Just upgraded to ChatGPT 4 but won't be renewing it.... it's a rip off that you pay $20 for 4.0 only to be limited to 50 messages every 3 hours 😡

Chat GPT rating you hot or not 😅

I wish you could do Vision Plus dalle and browsing all in one window so it keeps the same context
Without starting a new chat

Shit, Skynet is awake

I think what might help based on similar things I've done, that is, giving ChatGPT's own feedback from a different chat to itself, is probably copying and pasting the only the improvement suggestions you like the most and acting as if it's your own feedback. Then tell it to generate the same images again except including the minor suggestions this time. It might create less larger prompts immediately and not confuse the model.

I can't say for sure because I haven't gotten access to DALLE-3 yet to test it myself, but that's something I've done with things it's written for example, have a different instance analyze it and give suggestions on improvement, then paste the ones I like the most (if not all of them) back into the original chat.

you are abusing ppl attention it is way too long

Man how do you make money of this shit tho gang

Really, nonsense and futile video .just for view...not much outcome

Super cool stuff. Very exciting. 🚀

Personally I've been using it to to tune image prompts. I already utilize chat GPT to generate songs but the ability to given feedback on what actually is generated is huge. I also have been experimenting with using it to generate from to regenerate and image which it is not particularly great at but it is very interesting endeavor.
My technique varies a little from what you're doing in that I tell GPT to assume the image was generated from a prompt and then tell me what it thinks the proms would be to generate that image. I do also use GPT for image analysis but it is a different thing altogether and I generally get it to give to more technical terms as an artist when providing feedback on an image loading things like composition etc. Another thing I do is I tell it to get as close to 600 character count as possible because being image Creator is limited to roughly 600 characters but you have to make sure that you tell it to include spaces and punctuation as well as any tags or whatever it might include such as foreground background color make sure you tell it to not include the word pallid as well unless you want a grid of colors because if it sees the word pallet it will produce a pallet rather than using those colors as a basis four-color weight in the image.

Hellos. So, I finally got access to the Dall-E side of ChatGPT. I've been experimenting with prompts for ChatGPT and Bing Image Creator as well as picking it's cyber brain. I thought I'd pass this along from ChatGPT itself in case anyone else is noticing differences between image generations inside ChatGPT vs Bing Image Creator. (I asked Bing Chat to describe Image Creator and pasted its response into ChatGPT to fill it in since the Dall-E mode doesn't have internet access). "I'm integrated with a version of DALL·E based on the original model introduced by OpenAI. As of my last training cut-off in January 2022, I don't have explicit knowledge of subsequent versions such as DALL·E 2 or DALL·E 3. Therefore, my image generation capabilities are based on the original DALL·E model. If there have been updates or newer versions released after January 2022, I might not be directly integrated with them...Based on the information you provided earlier about Bing Image Creator being initially powered by DALL·E 2 and later integrating DALL·E 3 since March 2023, it suggests that Bing Image Creator is using a newer version of DALL·E than the one I'm currently integrated with. This could potentially mean that Bing Image Creator has access to more advanced or refined image generation capabilities compared to my current integration." What I've found to be the best workflow so far is to get ChatGPT/ Dall-E to generate the prompts and then paste them into Bing Chat. I'm wondering how long it will be before ChatGPT is trained on Dall-E 3?

how do i get access to DALL-E in chatGPT, i can't believe i've been subscribed to chatGPT for so long and i don't get any early access or nothin'? come on

I upgraded to GPT-4 today and the default doesn't have a button to add an image.
Isn't this automatically available? Are we waiting for the ability to add an image? I'm confused.

IMAGINE THIS: Instead of Photos, it understand Live video. All you need is a camera in a glasses and Voice feature in the glasses mini headphone and you have the ultimate assistant. WOOOOOOWWW!!!!!! 🤯🤯🤯🤯

That's why DALLE-3 is so precisive. There was no unnecessary/non-existent information like in Bard, so model can learn on really good data set (even more precisive than human can describe - humans are biased enough). Now Stability AI and other image generation projects can use GPT-4 vision to create a new great model. This is a very exciting technology indeed.

Ugh! STILL on 1.2023.270 and STILL no vision. I want it so BAD.

This is just beyond extraordinary. Really did not think I'd see AI like this in my lifetime. Great video, many thanks!

Does anyone know how to get Dall-E3 to generate consistent images? For example; an pink elephant named Lina in three different poses in the same setting.

👓

Didn't even notice the image button haha (I only use gpt when I need it) You're awesome dude! Now this will be my whole day

I have a premium account for chatgpt, but don't see the attachment button for uploading images.

Does anyone on here know why this is?

WE can't use it to identify people, but the people at OpenAI; and by extension the people at Microsoft, absolutely can. That's way more scary to me than if everyone at this capability.

hi matt i remember i used to watch u when u had around 10k subs lol
coming here after a long time man

i didn't see this button to upload image on chatGPT plus account, could you have any link to register to OpenAI for this part. Thanks.

Wow, all this hype and it can tell you that the bike has no wheels, spectacular. Waste of time

You do not use Chat GPT 5 already? Really cringe bro.😮

Too bad half of this is not accessible in the EU without VPN

Looking good MAtt!

any idea on context limit of the gpt-4V in browser? and how does images affect context limit? (any idea on how many tokens they count as if any ect?)

love your videos and find them really informative

GIMME!

why do I not have the option to upload images? Has it not been rolled out to everyone yet?