Quick Links
Having tried multiple AI image generation tools, I think that DALL-E 3 is up there with the best. It’s certainly not without its limitations, but in my opinion, it makes creating AI images from text incredibly straightforward. Here’s how I use the platform to do this.
1Find Inspiration From Other Creations
AI image generation is very different from photography. While I do look at other pictures for inspiration, I don’t do this very often—which is probably because I’m already quite skilled in that particular area. With AI image generation, I sometimes have an idea in my head but will also look at others for inspiration.
I tend to look specifically for AI-generated images to give me more ideas. However, I’ll also check out ordinary photos and illustrations to see if there’s anything that I can draw (pun intended) from these. Instagram is one place to look for this kind of content, but you can also check out different blogs.
If you’re in the same place, understanding variousDALL-E 3 prompts for your next imagecan also help.
2Start With Specific Prompts
Once I have a clear idea of what I want to create and know how to get there, I’ll start with specific prompts to bring my thoughts to life. Over the many months that I’ve experimented withDALL-E 3and otherAI image generators, I have slowly fine-tuned my prompts.
In the beginning, I did what every other beginner does; DALL-E received a basic prompt that I expected it to do wonders with, and it unsurprisingly didn’t give the results I was looking for.
These days, I’ll be very clear with the concept I want this tool to generate. For example, I used this prompt to make a picture within DALL-E 3:
Please generate a photorealistic image of a golden retriever in a Scandinavian-style home. I want the image to be set at Christmas time, and if you could give the dog a Santa hat, this would also be great. I do not want other people in the picture.
I have to say that I was very happy with this result because there have been multiple occasions in the past where ChatGPT has given me substandard results. In other cases, the software ignored the things I explicitly said not to feature. Here was the initial result:
3Ask for Adjustments if Needed
Sometimes, I’ll be happy with an image in DALL-E 3 and don’t need to ask the tool for further edits. The picture I generated above is one example. But despite this, it’s more common that I will need to ask DALL-E 3 for adjustments. For example, there have been multiple occasions where the app did not spell names correctly.
This doesn’t happen that often with images of animals. But if I want it to make a cityscape, issues often arise. In another example, I gave the AI the following prompt:
Please generate an image of a street scene in Paris on a rainy October day. I want there to be a café with writing on it
Admittedly, I wasn’t as clear as I should have been here. But as you’re able to see below, the writing on the café is blurry, and this was not my intention.
To try and tackle this problem, I asked ChatGPT to fix it:
Thanks for creating this, but I cannot read the text on the café building. Can you please change the text to say Le Café Parisienne and put it all in capital letters?
The reason I asked DALL-E 3 to amend the image was because it didn’t detect text when using the different editing tools; this is one of many reasons thatDALL-E 3’s integrated editing features still need work. As you may see, the text was more readable but still not 100%. So, you’ll have to experiment because this often takes some time to get right.
4Edit Specific Parts of the Image
While DALL-E 3’s editing tools are far from perfect, you might still find them useful when you don’t want to make full-scale adjustments. I would recommend using these just for blemishes. To do this, I’ll go back to the initial image of the dog in a hat.
First, you’ll need to select theedit icon(it’s the palette and brush).
After doing this, I then used the circular brush to edit a specific part of the picture. I wanted to make the left side of the background blurrier, and so, I asked DALL-E 3 to give me more of an out-of-focus effect: “Please make this section more out of focus to represent a bokeh effect.”
The background was slightly more blurry, but it’s still not what I wanted. Again, this was another case of not being specific enough. In hindsight, I’d tell DALL-E 3 to pick a specific focal length (e.g. f/2).
5Try Different Image Prototypes
As you’ve seen so far, DALL-E 3 isn’t always going to deliver perfect instant results. Because of this, I suggest creating different image prototypes. In my experiment, I asked DALL-E 3 to regenerate this image of the dog on more than one occasion. Because I didn’t get the full background blurriness I had tried with editing, I asked DALL-E to do this in the normal chat.
The dog wasn’t smiling in the picture it gave me, so I then asked DALL-E to add a smile to the dog. This was the result:
As you may see, the background is blurrier than it previously was. I would then go in and make more image types, before picking the one I think best suited my needs.
6Resize the Image Elsewhere
While I’m happy to try editing different aspects of an image in DALL-E 3, I never resize within the app. I’ve tried 20+ times in the past, but the tool just gives me the same dimensions and it becomes frustrating after a certain point. Instead, I’ll download the picture and resize it from my computer.
Generally speaking, I think thatCanvais the best tool to resize images. I normally pick my dimensions and then set the picture as a background. From here, I can crop to include or exclude certain elements. You can also use your smartphone’s Photos app (or equivalent).
It’s possible to create all kinds of images in DALL-E 3, though it will take some time to get everything right. I recommend making multiple types of the same image, and you can also try using the editing tools.
