Oct 13, 2023·edited Oct 13, 2023Liked by Daniel Nest
I do appreciate the DALL-E 3 integration, but I'm still working on re-training my muscle memory since I'm so used to opening up Midjourney. One thing to try is having GPT-V describe a DALL-E image, then having DALL-E regenerate it, etc. AI pictionary!
Tell me about it. I'm so used to typing "/" and having Midjourney automatically complete it to "/imagine," then pressing Enter to type in my prompt that I've done this on autopilot within ChatGPT DALL-E 3 several times. Not even joking: https://i.imgur.com/nv3UYf4.png
And yes: That's exactly what I'm imagining when all the separate "Beta" features are integrated into a single chat within ChatGPT. Browse for a PDF link and have Advanced Data Analysis review it. Feed GPT-V an image from DALL-E 3 (or anything else) and have it reimagined, etc.
You can sort of do this today with the "/describe" command with Midjourney and its own images. It usually works quite well, and since it generates 4 alternative prompts, you always end up diverging from the original ever so slightly in unpredictable ways.
I have had a little bit of time to play with some of the features, and I don't think they're quite ready for prime time. I'm sure I'd feel really differently if Bing wasn't already my go-to image generator (better quality than GPT so far from what I can tell, but perhaps able to close that gap), or if I wasn't inclined to do research with Bard (it's good at factual quick stuff, and hallucinates a lot less than GPT4, at least for what I do). As it stands, I'm glad to keep using those services, at least for now, although using just one fully integrated window would be fantastic. I just don't think that even the integrated stuff in one chat will make it worth exclusive GPT4 use... but over time, that might change.
My experience with Bing DALL-E 3 vs. ChatGPT DALL-E 3 has been the exact opposite: I find the ChatGPT version to return more polished and sharp images. If you come across some specific examples, I'd love to see them.
My go-to image generator is still Midjourney if image quality is what I'm after. But DALL-E 3 is definitely the official cartoonist for AI Jest Daily. I start out with ChatGPT but sometimes end up going to Bing because of the "seed" stuff I talked about above.
Roger that! I've been kind of reading what you suggest and running over and trying it, but I'm starting to understand that I might have to wait to play with these toys. I'm also probably conflating all the different attempts Open AI is integrating, like Code Interpreter, etc - little things they keep sort of trying or discarding, so I mix up what the app is capable of.
Ultimately, if I can use it every day, I'll really start to understand and perhaps "master" the image generator, whatever it is, but I'm also pretty happy with the other options I have for now, so that's cool too. I will continue to follow your lead on reaching out and experimenting.
Excellent post! Thanks for writing it.
Thank you, I'm happy you found it useful!
I do appreciate the DALL-E 3 integration, but I'm still working on re-training my muscle memory since I'm so used to opening up Midjourney. One thing to try is having GPT-V describe a DALL-E image, then having DALL-E regenerate it, etc. AI pictionary!
Tell me about it. I'm so used to typing "/" and having Midjourney automatically complete it to "/imagine," then pressing Enter to type in my prompt that I've done this on autopilot within ChatGPT DALL-E 3 several times. Not even joking: https://i.imgur.com/nv3UYf4.png
And yes: That's exactly what I'm imagining when all the separate "Beta" features are integrated into a single chat within ChatGPT. Browse for a PDF link and have Advanced Data Analysis review it. Feed GPT-V an image from DALL-E 3 (or anything else) and have it reimagined, etc.
You can sort of do this today with the "/describe" command with Midjourney and its own images. It usually works quite well, and since it generates 4 alternative prompts, you always end up diverging from the original ever so slightly in unpredictable ways.
Great article! Thank you.
You got it! Happy you found it useful - don't hesitate to ask if you have questions about any of the features.
Thank you. 🙂
I have had a little bit of time to play with some of the features, and I don't think they're quite ready for prime time. I'm sure I'd feel really differently if Bing wasn't already my go-to image generator (better quality than GPT so far from what I can tell, but perhaps able to close that gap), or if I wasn't inclined to do research with Bard (it's good at factual quick stuff, and hallucinates a lot less than GPT4, at least for what I do). As it stands, I'm glad to keep using those services, at least for now, although using just one fully integrated window would be fantastic. I just don't think that even the integrated stuff in one chat will make it worth exclusive GPT4 use... but over time, that might change.
My experience with Bing DALL-E 3 vs. ChatGPT DALL-E 3 has been the exact opposite: I find the ChatGPT version to return more polished and sharp images. If you come across some specific examples, I'd love to see them.
My go-to image generator is still Midjourney if image quality is what I'm after. But DALL-E 3 is definitely the official cartoonist for AI Jest Daily. I start out with ChatGPT but sometimes end up going to Bing because of the "seed" stuff I talked about above.
Okay, you have me convinced to create images inside of ChatGPT for a few days in a row now.
Curious to hear your thoughts and how you feel it compares to Bing once you have a verdict.
Dan, are you able to generate images directly from ChatGPT 4? I do not have the Dall E option on my dropdown menu yet. Maybe I need to wait.
Correct: I am able to do exactly that. That's what my entire #1 section is based on, along with the screenshots. Hope you get it soon as well!
Roger that! I've been kind of reading what you suggest and running over and trying it, but I'm starting to understand that I might have to wait to play with these toys. I'm also probably conflating all the different attempts Open AI is integrating, like Code Interpreter, etc - little things they keep sort of trying or discarding, so I mix up what the app is capable of.
Ultimately, if I can use it every day, I'll really start to understand and perhaps "master" the image generator, whatever it is, but I'm also pretty happy with the other options I have for now, so that's cool too. I will continue to follow your lead on reaching out and experimenting.