DALL-E 3 Tips & Tricks: Maximize Your AI Art Skills with ChatGPT-4!

in this video I will demonstrate some tips and tricks for using Dolly 3 with chat gp4 to use Dolly you'll require A Plus subscription and must utilize version 4 as version 3.5 doesn't support Dolly it's essential to note that you're restricted to 30 messages per 3 hours or if you have the team version you'll have a limit of 100 messages per 3 hours you simply prompt for what you want to create in this case a cartoon bunny with an egg in a watercolor painting style now if you're satisfied with what you received and wish to create a new one with a green egg you can specify in the prompt that you want to use the same seed now let's try another one with the same seed but this time let's make the egg blue as you can see the blue and green ones are almost the same character with only the egg being different however when I prompt for one with an orange egg it's slightly different but still very similar to the first one so with the same seed you get a slight variation of the image sometimes they're very similar but other times not so much uh let's delve deeper into understanding the seed after generating the first image I'll edit my prompt and ask for the seed and image uh generation ID chat GPT will then provide me with some text the generation ID is composed of both text and numbers uh while the C consists only of numbers for instance if you have these numbers you can use them to refer to specific images here I attempted to recreate the same image using those numbers but it was one of those moments when chat GPT got lost so I tried again using the word regenerate and I was able to regenerate the exact same image with the generation ID you can refer to any image you generate but only within this session therefore you can go back find your generation ID and refer to that image when you want changes for example in this case I use the same generation ID to get a purple one instead of red I understand that this may seem confusing at first so let's clarify it a bit to help you better understand the generation ID gen ID stands for Generation identifier and is a unique identifier assigned to each image generated by Deli 3 it's used to reference a specific image that was created if you want to create variations uh or or further iterations of that image you can use the Gen ID to specify which image you're referring to think of it like having a specific file name for each image if you tell Dolly I want to modify the image with Gen ID xyz123 it knows exactly which image you're talking about this makes it easier to manage and work with specific images within your generated content a seed is a starting point for the random number generator used in the ai's creative process the seed determines the randomness of the image generation if you use the same seed for different prompts you may notice similarities in style or patterns because the ai's decision-making starts from um the same point think of it like planting a tree if you plant seeds from the same Apple you'll get trees that are similar to each other in Dolly's case using the same seed for different images can bring a degree of consistency in uh style or feel here are some key differences gen ID is about referencing a specific image already created seed is about where to start the creative process each gen ID is unique to an image seeds can be reused to generate images with similar Styles or patterns gen ID is used after an image is created while a seed is used before or during the image creation process Let's test it with another example I will create a cute cartoon dog then if I want to change the ratio of the image I'll ask to keep the same generation ID and to make it a wide ratio you can do the same to make Square landscape or portrait ratios it's a flipped version in landscape mode but it's quite similar let's try the same but this time with a portrait ratio it's very similar so that generation ID can help you have a little more control over consistency let me demonstrate my approach to a project first I start with a prompt giving it a basic idea of what I want then I examine the image and decide what changes should be made I interact with it as I would with another human using the same generation ID to communicate that I want something as close as possible to that specific image if I desire a completely different image I simply regenerate for instance here I obtained a similar painting with a red hat while it's not perfect the POS and colors are similar with slight changes I could even go further and create a winter scene with a Santa hat forming a series of images in a similar Style for a collection or set additionally I can add text though maintaining the exact same image with text is challenging however it remains within the realm of style and composition when prompting for an image you can include more than just objects and characters you can also include mood emotions and art styles to help the AI understand better what you want to create for instance as I did here you can request an impasto painting of a sad cat in the rain set in a steampunk City this provides the AI with a clearer understanding of the desired image incorporating specific elements such as texture emotion and setting to create a more nuanced and tailored result here I prompted for a night um typically associated with the past but set in the future and in a vector style this approach allows for the combination of elements not commonly found together resulting in unique and interesting Generations uh it knows all kinds of art styles so you can choose styles that fit your mood and idea better for example here I requested a pencil drawing of a hand drawing another hand who said that AI is not good with hands haha well it's good but it's not perfect and definitely Dolly makes fewer mistakes than uh other generators let's try something cute in watercolor style as you can see in the prompt I included the subject which is wearing a crown the environment where it's sitting the adorable look on its face the emotion it conveys love and the mood and lighting morning light rays finally I specified the art style as watercolor painting you can add more details if you want and such as describing the texture of the paper or the frog's skin the AI has its limits in terms of what it can keep track of but usually one or two sentences are enough for it to understand here here I was able to achieve a similar look without using the generation ID or seed first I expressed that I like the initial result prompting the AI to attempt to generate a similar version then I specified what to replace in the image such as the color of the Frog the AI then tried to understand my preferences and maintain them within a certain range when creating the variation if I had requested another one it would have provided a completely different image because it was attempting to discern uh my preferences and and generate different versions until I found one I liked then I asked for another change requesting the Frog to hold a heart however I didn't like that it cut too much from the side so I used the regenerate button the resulting image wasn't identical but it was quite similar especially considering that I was guiding the process through verbal instructions much like how one would communicate with an artist and direct them on what to paint or change change then I wanted to generate a wide ratio a landscape mode image perhaps for use as a YouTube thumbnail for this video as you can see I also wanted to include a sign with text on it however the first generation misspelled the word di so I regenerated to get another version with the correct spelling if you don't want to spend too much time prompting you can explore some GPT models that other people have made by clicking on explore GPT and then selecting the doll e tab you can find a few suggestions here or you can search for all public GPT models I've tested a few and let me show you a couple that I found interesting one of them is super described which allows you to upload an image and it will try to generate a similar variation of that image as you can see the results are quite similar additionally if you click on the info button you can see the actual prompt and description which can be useful for understanding how it works let's click again on explore gpts and this time select the consistent character GPT here they've made it easy you just click where it says click here to start it will ask you if you want a male or female character then it will prompt you for a name and more details about the character let's go with something wild like a a black ninja bunny with red eyes when it asks you for style it gives you four options but you can also use your own style like a watercolor painting cartoon style from there you can play with emotions poses and all kinds of things and should get similar characters the first one might not be so similar but the second third and fourth ones should be quite similar you can then add environments change all kinds of things to make it look how you need it if you found something helpful in this video take a second from your time and hit the like button it's free and will help me make more tutorials for you thank you for [Music] watching

Channel: pixaroma

Views: 1,489

Keywords: dall-e 3 chatgpt, dall-e prompts tips, dall-e image generator, dall-e ai, dalle 3, dalle 3 chatgpt, dall e 3 chatgpt 4, dall e 3 chatgpt prompts, create images in chatgpt with dall-e 3, chatgpt 4.0 dall e 3, dall e 3 tutorial, dall-e 3 tutorial, dall-e 3 tips and tricks, how to use dalle 3, dall e 3, dall-e 3, dall-e 3 prompts, dalle 3 gen id, dall-e 3 gen id, dalle 3 consistent characters, consistent characters using dalle 3, create consistent characters ai, pixaroma

Id: gcNBCbuW9jg

Length: 10min 14sec (614 seconds)

Published: Tue Feb 13 2024