If you use a lot of chatgpt, you have probably experienced Dall-E 3, the last OpenAi iteration of an IA images generator. Dall-E 3 can be shocking in a shocking way to give life to your ideas, but sometimes badly shocking to understand certain details or simply shocking what he chooses to emphasize. However, if you want to see a specific scene moving from your head to your screen, Dall-E 3 is generally quite useful, it can even have your hands write.
But here is the thing, Dall-E 3 is always a model of AI, not a reader of mind. If you want images that really look like what you imagine, you need to learn to speak your language.
After a few tests and errors (and some involuntary horrors), I became quite skilful to speak his language. Here are five key tips that will help you do the same.
Polon
Dall-E 3 has a quality mode called “HD”, which makes the images sharper and more detailed. Think about it as the addition of a high -end camera lens to your art generated by AI – textures are richer, lighting is more refined, and things like fabric, fur and reflections have more depth.
Find out what it looks like in this prompt: “A rendering of a close plan of a butterfly based on a sunflower. Quality: HD.”
Play with an appearance
Not all images should be a square. Dall-E 3 allows you to define the aspect ratios, which may seem minor but can be huge if you want to make something more cinematographic or as a portrait.
Square is fine, but why limit you when you can do something epic? This is particularly useful for graphics and posters of social media, such as the one below, which had the prompt: “A vertical poster of a vintage travel ad for Paris, size 1024×1792 pixels.”
Think like a director
To get an image capable of evoking emotion, it sometimes helps to think as if you were a photographer in the real world. Think of the angle of the camera or composition techniques; Look at them if necessary. The result can considerably change the appearance of an image.
Instead of a flat and dead image, you can ask for angles like the close -up, the bird’s view or the excessive shoulders. The same goes for styles of composition and terms as “symmetrical composition” or “field depth”.
This is how you can get the following image from this prompt: “A dramatic photo in the shoulder of a lonely cowboy standing on a rocky cliff, looking at the large landscape of the desert below. The sun is Laying in the distance, throwing long shadows through the canyon.
Iter, iter, iterer
One of the less known but very effective tips of Dall-E 3 is to tell him what to include. This avoids unwanted elements of your image. This could mean specifying negative elements such as colors, objects or styles that you do not want or do not refine style and mood by what you don’t want it to look like.
This is how I got the image below, using the prompt: “A peaceful park in autumn with a young woman sitting on a wooden bench, reading a book. Golden leaves cover the ground And a soft breeze brushes the trees.
Be too precise
Consider Dall-E 3 as a very literal genius: this gives you exactly what you ask, no more, no less. So, if you type “a dog”, do not be surprised when it spits a random dog of race, atmosphere or indeterminate moral alignment. The more details you include – such as breed, color, decor, mood or even artistic style – the better the results.
For example, you could start with: “A sorcerer throwing a spell”, but you would better submit: “A sorcerer elderly with a long braided white beard, dressed in emerald green dresses embroidered with golden runes, evoking a vortex swirling blue lightning with the tips of your fingertips in a stormy mountain landscape. You can see the two below.
