Unfortunately the Custom instructions I wrote became somewhat defunct after an update to ChatGPT, preventing you from generating multiple prompts at once. So instead I implemented all the same features, in a custom GPT. You can try it here: chat.openai.com/g/g-7CKojumSX-the-glibatree-art-designer
Your video just made me decide to get ChatGPT. So, I have a lot of learning to do. I hope to see more tutorials coming from this channel. I hit sub already.
Sir. You have my like, subscription, discord join, and loyalty. I am utterly astounded at these results. You, Sir, are an answer to prayer. Super duper quadruper THRILLED. Thank you times a Kuhzillion for these helpful hints/hacks. I am FLOORED. 😮 - Randall
This is amazing! The results are absolutely stunning! I can’t thank you enough 😊 On a side note, I find that this works beautifully with very very simple prompts, therefore letting DallE imagine whatever it wants. But when you want something very specific and you decide to write quite a long detailed prompt of the picture you have in mind, the script doesn’t help and doesn’t add any additional details.
For long prompts, I would say give ChatGPT an opening... Say something like, "I'm not sure how to make the lighting work, so while writing the prompt please be sure to describe how it interacts with each item" My prompt engineering does two things: 1. It makes sure the prompt isn't underdefined, so basic ideas can create great images 2. It gives the GPT-4 a prompt format, so it can add a lot of detail without confusing DALLE-3 Before my custom instructions, I found the more intense my image was, the more likely cameras and lighting gear or paint brushes would find their way into the actual picture. This fixes that. Hope that all makes sense, and be sure to join my discord, i'd love to see what you come up with 😃
@@glibatree That’s very helpful, thanks! It actually made me understand more how DallE works. I would never have thought I could give it instructions on the camera aperture or shutter speed 😮 I’ll definitely experiment a lot more both with your script and without. But one thing for sure, images with the script are crazy realistic and/or detailed. Way more that what I achieved with very detailed prompts. I’m not very familiar with discord but I will give it a shot! 🫡
I really appreciate you sharing this information. Your images are quite impressive and I too liked and submitted. I'm using DALL-E and others as a base for my interior design business, so I will see if I can get my images to look better. Thanks again!!😊❤
Thanks for this. I just started messing with Dalle-3 in Chatgpt yesterday & was already impressed. I can't wait to see what the same prompts will look like now.
Thanks for the video. Basically it also works, but in the last few days images created with custom instructions are very often simply not created and an error message appears. Presumably Open AI has implemented strict restrictions here to prevent custom instructions for image generation. If you switch these off, you get significantly more images and without error messages.
In my experiments I found the custom instructions are mostly just more prompts. I call it preprompting but it would work pretty much the same as if you typed it out each time.I use it to experiment with interactive fiction. For dalle 3 usage it will forget older prompts after a while.
Amazing stuff, thank you very much for this; Can I sum to you amazing custom instructions. Ask to always give you the seed number of each image generated. Like that you are able to edit the image you like etc.
There actually is a slight difference to the interface when you DALL-E 3, which is that the little square in the chat box, allowing you to update images for ChatGPT to analyze goes away. It’s interesting that it will either create images, or analyze them, but not both (in the same session).
OK, I gave it a whirl On small issue, Dall•e 3 has a letter count issue, so you need to limit the number of words used, or characters, but it was easy enough to sort that out and add as an additional item in the custom requirements list. Thanks again. Was useful.
I watched this video 2 weeks ago and have been getting amazing results, thank you. Saw a video just posted from another creator giving credit to someone else for coming up with a very similar prompt to this. Bit disappointing, but I guess once you say stuff on the internet it’s open for copying.
Agreed yeah, once something is out there it kind of finds its way everywhere - and sometimes it is hard to trace back to where it started. But, at the same time, I think a rising tide lifts all ships. So the more the community figures out the better our generated art can become. So my prompt gets someone else thinking about how they would write their own, and that one becomes more popular or whatever, it's all for the best. The ecosystem of prompts shall flourish haha.
Thank you so much for sharing such a great video. I tested it myself and definitely saw a huge difference. I endorsed your channel on mine. Thank you again.
been trying out all the different a.i and so far dalle3 has been giving me the best results despite not having a upscale to increase the size, dont know if theres a way to resolve this so that it increases the quality. The more popular dalle3 gets the more likely it will end up behind a paid wall to use it.
AI for art is by far the most interesting thing I've come across in my lifetime. It's almost akin to magic in my mind. It completely delegitimizes art as an endeavor, though. I question everything I see and am on the verge of completely abandoning by ambitious to be a better artist.
For anyone that might be confused, you can make your own GPT with the the files in the bin link. I just did and effectively the result is identical, although it does give you more room to adjust it even further if you like to feel unique and special ;)
Not often I do that, honestly, words "like and subscribe" are more like a noise nowadays you just get used and don`t hear it, ut dude, you`ve got mines.
The paste bin top section is 4 characters too long (3 after I removed an errant whitespace) I found the rest, there were 4 errant white spaces making the text 4 characters over 1500.
Yeah, you're not the first person to say you had to clean it up, I'll see if I can do the same inside of Pastebin so others don't have to deal with this. Good catch, thanks 🤩
These are cool images, but they don't look like photographs, like Mid-Journey can create. Are there any prompts you can recommend to get the image to be more realistic?
Dalle3 can make everything Midjourney can make & is integrated with ChatGPT. I’m not sure what you’re doing wrong if yours isn’t working. Just ask it for a photograph
I’m trying to work out if order etc matters eg for Leonardo this works quite well: # Constructing the main prompt main_prompt = ( f"{topic} {image_type} {artist}. " f"The room features {elements}. " f"It radiates a feeling of {feeling}. " f"The wall is painted in {color_scheme}. " f"The room includes textures and patterns like {textures_patterns}. " f"The furniture is made of {furniture_wood}. " f"The room has {lighting_mood}, " f"set in {building_environment}." ) So far I find copying my Leonardo prompts in to start with works better but then the advantage is chatting about what I like and don’t like and revising the images. Sometimes though it looks less photorealistic than Leonardo- I think things are maybe too crisp and shiny?
YES! Love the way Dall-E 3 is looking, but realistic faces are still the best in MidJourney. If Dall-E could do that, and brought inpainting into it, it would be my number 1!
This is a very nice tool. It keeps putting eyeglasses on my characters even though I tell them to make it historically accurate. Any ideas on how to get it to quit doing that? Thanks.
If you select raw mode in the paste bin, it's under 1500. Paste bin seems to add some extra spaces on new lines which is why it fits over. You can just delete those manually if that's easier
It has changed for me recently. It only allows one or two images. When I ask it to create more, it says it can not generate more. It says I have to change to prompt.
I didn't work for me. I copy/pasted the exact prompt and got the same images I had been getting before using custom instructions. Back to the drawing board.
I wish Chat GPT 4 and Dally 3 allowed for picture upload so I could upload picture for Dally to get inspired from or extract some element into my new image.
Unfortunately the Custom instructions I wrote became somewhat defunct after an update to ChatGPT, preventing you from generating multiple prompts at once.
So instead I implemented all the same features, in a custom GPT. You can try it here:
chat.openai.com/g/g-7CKojumSX-the-glibatree-art-designer
And this is the video from testing that one out:
th-cam.com/video/xyTwyKBWxLM/w-d-xo.html
so is this private? will my images stay private using your page?
@@glibatree
I smacked your Like button and hit subscribe to boot..
Watching from Australia..
Truly amazing! Thank you!
Your video just made me decide to get ChatGPT. So, I have a lot of learning to do. I hope to see more tutorials coming from this channel. I hit sub already.
This has given me so much, thank you!
Wow thank you! I am honored 🤩
Wow, that custom GPT is off the hook! Very cool! I'm trying to use chatgpt for my thumbnail creations and this will help tremendously! I subbed!
Great prompt, really helpful to improve the Dalle 3 image quality! Thanks for sharing!
Sir. You have my like, subscription, discord join, and loyalty. I am utterly astounded at these results. You, Sir, are an answer to prayer. Super duper quadruper THRILLED. Thank you times a Kuhzillion for these helpful hints/hacks. I am FLOORED. 😮 - Randall
You offered something, said it was free, and it actually was. Liked and subscribed!
you can also paste that full script into the prompt window along with the desired image prompt and it works as well.
Very cool! The castle images are so inspiring.
Right! Especially first painting with the swirling brush strokes... Gorgeous
I'm so happy I fell on this video, I had no ideo of all these features. Thanks buddy
omg, so much fun
Thank you so much!
enjoy a cup of good coffee @@glibatree
I watched and skimmed this video, and now I'm back just to say thank you. Your shares are genuinely helpful. Once again, I sincerely thank you. ❤
Thank you! This has seemed to fix the issue I had with my images coming out portrait shaped but with everything in the image sideways.
That is super obnoxious. Sometimes I ask it to 'rotate those image 90-degree to the right'
This is amazing! The results are absolutely stunning! I can’t thank you enough 😊 On a side note, I find that this works beautifully with very very simple prompts, therefore letting DallE imagine whatever it wants. But when you want something very specific and you decide to write quite a long detailed prompt of the picture you have in mind, the script doesn’t help and doesn’t add any additional details.
For long prompts, I would say give ChatGPT an opening...
Say something like, "I'm not sure how to make the lighting work, so while writing the prompt please be sure to describe how it interacts with each item"
My prompt engineering does two things:
1. It makes sure the prompt isn't underdefined, so basic ideas can create great images
2. It gives the GPT-4 a prompt format, so it can add a lot of detail without confusing DALLE-3
Before my custom instructions, I found the more intense my image was, the more likely cameras and lighting gear or paint brushes would find their way into the actual picture. This fixes that.
Hope that all makes sense, and be sure to join my discord, i'd love to see what you come up with 😃
@@glibatree That’s very helpful, thanks! It actually made me understand more how DallE works. I would never have thought I could give it instructions on the camera aperture or shutter speed 😮 I’ll definitely experiment a lot more both with your script and without. But one thing for sure, images with the script are crazy realistic and/or detailed. Way more that what I achieved with very detailed prompts. I’m not very familiar with discord but I will give it a shot! 🫡
I get insane results and I can get exactly what I have in mind. (Except for a few errors sometimes of course, it’s still AI 😆) Thank you so much!
@@Milky63 awesome news! Keep it up 💪
you have gained a like good sir, I applied this to my gpt and it works great, thank you
This is really great stuff, thanks for sharing!!
Can't wait to try this! Thanks for sharing!!!
I am blown away! Thank you so much for this tutorial. Thank you for being so thorough and for giving us the prompts! ❤❤❤
curious what everyone does with these pictures.
I really appreciate you sharing this information. Your images are quite impressive and I too liked and submitted. I'm using DALL-E and others as a base for my interior design business, so I will see if I can get my images to look better. Thanks again!!😊❤
Wow, the results are spectacular! Liked and subscribed.
Immediately subscribed after finishing this video. Appreciate you. Thank you brother
I also used custom instructions and was satisfied, but yours is gold!
Thank you, sub, like and I'm waiting for the next videos. Incredible.
Thank you very much! You're very kind!!
Boy you're really cool, thank YOU SO MUCH, I'm already using these instructions and it's epic
Thanks for this. I just started messing with Dalle-3 in Chatgpt yesterday & was already impressed. I can't wait to see what the same prompts will look like now.
Good luck, I'd be curious to hear about the results!
Thanks for the video. Basically it also works, but in the last few days images created with custom instructions are very often simply not created and an error message appears. Presumably Open AI has implemented strict restrictions here to prevent custom instructions for image generation. If you switch these off, you get significantly more images and without error messages.
Im also getting errors now. Thats too bad
You are my hero! Liked, subscribed, and hit that bell! 👍🏽
Thank you for posting this. It's awesome. I subscribed and look forward to seeing more of your videos.
You have a new subscriber I used your prompt and amazing results thank you
Since I use my special instructions already, I tried making a GPT from this. I am happy with my results! Thanks a lot! Oh, I see you did the same :D
OMG! This is such a game changer. Thank you for sharing.
Like button smacked, thanks!
Very very good
This is an amazing video. Thanks for sharing!
Thank you so much. This is absolutely amazing!!!
I tried it and it works like a charm. LOVE IT
That's awesome dude, keep it up!
You are a genius at telling robots what to do.
Insane results. Wow! Thanks a lot, following your new content!
I don’t know who this dude is, but he’s likable AF…subbed
Thanks for making a free version of your method!
Thank you! Priceless! Ale the best to you ;)
Lmao the way you opened this video up was funny!
thank you so so much for this! subbed!
were not worthy were not worthy! you are a god
In my experiments I found the custom instructions are mostly just more prompts. I call it preprompting but it would work pretty much the same as if you typed it out each time.I use it to experiment with interactive fiction. For dalle 3 usage it will forget older prompts after a while.
Liked + Subscribed ❤
Will try to use these techniques for my ferrets images
Thanks for the prompt. The photos are much better.
Awesome man! Hope you enjoy
Your GPT is fantastic, its my favourite, thanks.
Wow. This is sooooo helpful!!
Amazing stuff, thank you very much for this; Can I sum to you amazing custom instructions. Ask to always give you the seed number of each image generated. Like that you are able to edit the image you like etc.
There actually is a slight difference to the interface when you DALL-E 3, which is that the little square in the chat box, allowing you to update images for ChatGPT to analyze goes away. It’s interesting that it will either create images, or analyze them, but not both (in the same session).
Oh yes! Good point, that is interesting haha
I wonder if it can do infographics@@glibatree
OK, I gave it a whirl
On small issue, Dall•e 3 has a letter count issue, so you need to limit the number of words used, or characters, but it was easy enough to sort that out and add as an additional item in the custom requirements list. Thanks again. Was useful.
This is awesome. Now please find out a way to generate consistent images with DALLE-3 :)
Thank you so much for this ❤
Ty this is amazing, bro!
Thank you!!! Liked and Subscribed ❤❤❤
You can pick more useful aspect ratios like 16:9 and 9:16, which is what almost all social media uses primarily.
Outstanding!👏👏👏
Great work! thanks for sharing
Thanks man! I hope it gets you some good results~
I watched this video 2 weeks ago and have been getting amazing results, thank you. Saw a video just posted from another creator giving credit to someone else for coming up with a very similar prompt to this. Bit disappointing, but I guess once you say stuff on the internet it’s open for copying.
Agreed yeah, once something is out there it kind of finds its way everywhere - and sometimes it is hard to trace back to where it started. But, at the same time, I think a rising tide lifts all ships. So the more the community figures out the better our generated art can become. So my prompt gets someone else thinking about how they would write their own, and that one becomes more popular or whatever, it's all for the best. The ecosystem of prompts shall flourish haha.
Thank you so much for sharing such a great video. I tested it myself and definitely saw a huge difference. I endorsed your channel on mine. Thank you again.
This video is gold!
Hey, thanks so much!
Always amazing
great info. thank you
Worth it ❤
You are a genius
been trying out all the different a.i and so far dalle3 has been giving me the best results despite not having a upscale to increase the size, dont know if theres a way to resolve this so that it increases the quality. The more popular dalle3 gets the more likely it will end up behind a paid wall to use it.
You the best.
Got a sub from me. Clear and concise instructions!! Thank You
AI for art is by far the most interesting thing I've come across in my lifetime. It's almost akin to magic in my mind. It completely delegitimizes art as an endeavor, though. I question everything I see and am on the verge of completely abandoning by ambitious to be a better artist.
For anyone that might be confused, you can make your own GPT with the the files in the bin link. I just did and effectively the result is identical, although it does give you more room to adjust it even further if you like to feel unique and special ;)
Looking forward to more videos like this
Oh good haha! I'll see if I can deliver!
Not often I do that, honestly, words "like and subscribe" are more like a noise nowadays you just get used and don`t hear it, ut dude, you`ve got mines.
Instantly leveled up my generation game! Thanks :)
If this isn't a reason to subscribe, I dont know what is.
Wow! Extremely helpful and I can't wait to see what 's next. Thanks!
As of today it only generates 1 pic at a time! Very upsetting!
Very helpful thank you
They definitely need to give the users custom instructions per chat type, or selectable templates.
That would be a great feature for my workflows!
That’s a good idea. I’ve been doing my own, but there’s no reason they couldn’t do some for you.
The paste bin top section is 4 characters too long (3 after I removed an errant whitespace) I found the rest, there were 4 errant white spaces making the text 4 characters over 1500.
Yeah, you're not the first person to say you had to clean it up, I'll see if I can do the same inside of Pastebin so others don't have to deal with this.
Good catch, thanks 🤩
I had the same issue. Looking for the errant white spaces now...
Same here.. having issues finding them
Thank you sir!
I wish I could have recorded my screen but literally when you said slap the like button, the like button had a ring pop up around it.
These are cool images, but they don't look like photographs, like Mid-Journey can create. Are there any prompts you can recommend to get the image to be more realistic?
Dalle3 can make everything Midjourney can make & is integrated with ChatGPT. I’m not sure what you’re doing wrong if yours isn’t working. Just ask it for a photograph
THANK YOU !!!
I’m trying to work out if order etc matters eg for Leonardo this works quite well: # Constructing the main prompt
main_prompt = (
f"{topic} {image_type} {artist}. "
f"The room features {elements}. "
f"It radiates a feeling of {feeling}. "
f"The wall is painted in {color_scheme}. "
f"The room includes textures and patterns like {textures_patterns}. "
f"The furniture is made of {furniture_wood}. "
f"The room has {lighting_mood}, "
f"set in {building_environment}."
)
So far I find copying my Leonardo prompts in to start with works better but then the advantage is chatting about what I like and don’t like and revising the images. Sometimes though it looks less photorealistic than Leonardo- I think things are maybe too crisp and shiny?
Complete breaks chatGPT for me: Error Messgae: something went wrong... oh well...
Great video, thanks!
Wonder if there might be a way to guide dall-e 3 to make more realistic faces?
YES! Love the way Dall-E 3 is looking, but realistic faces are still the best in MidJourney. If Dall-E could do that, and brought inpainting into it, it would be my number 1!
The last job we had was to tell the machine what to create for us, now we leave that to one too.
This is a very nice tool. It keeps putting eyeglasses on my characters even though I tell them to make it historically accurate. Any ideas on how to get it to quit doing that? Thanks.
Very nice sir,,,plz tell me prompt command if I want to make realistic pic of any body builder or any actor,,,,
Thank you!
If you tell it to only so one picture the prompts are mich larger and the pictures are much better in quality also tell it to dk wide aspect ratio
Custom Instractions not working with the new version of GPT DALL-E 3 :(
how come my chatgpt only created 2 images, when i gave the custom instructions to generate four? am i missing something? any advise?
Unfortunately, the first box did not work. It goes over 1500 words.
If you select raw mode in the paste bin, it's under 1500. Paste bin seems to add some extra spaces on new lines which is why it fits over. You can just delete those manually if that's easier
It has changed for me recently. It only allows one or two images. When I ask it to create more, it says it can not generate more. It says I have to change to prompt.
Omg, thank you so much, you saved me. Let me know if you need anything LOL
WOW thanks man.
I use chat gtp 4 through the bing app. How do I input custom instruction through the bing app?
weird, I got the custom instructions in. But the prompts are not showing.
I didn't work for me. I copy/pasted the exact prompt and got the same images I had been getting before using custom instructions. Back to the drawing board.
I wish Chat GPT 4 and Dally 3 allowed for picture upload so I could upload picture for Dally to get inspired from or extract some element into my new image.
If have an image of a house and I want Dalle 3 to create an image of that same house but make it pop like that burger photo ... how do I do that ?
Tried it - everything I could think of was restricted so I couldn't make it.