AI Art Generation has evolved at a wild rate, and Google simply threw another big competitor in the mixture via its Gemini Flash 2.0. You can play with the new image creation tool in the Google AI studio.
Gemini Flash is, as its name suggests, very fast, especially faster than Dall-E 3 and other image creators. This speed can mean lower quality images, but this is not the case here, in particular because all changes and upgrades of the model image production capacity. However, if you want very good results, you should know how to talk to AI. After a lot of tests and errors, I set up five tips to withdraw the best absolute art from Gemini Flash 2.0. Some of them may seem similar to advice on other art creators, because they are, but that makes them less useful in this context.
Tell a story
The new most interesting feature for the creation of images of Gemini Flash is that it is not only good for occasional illustrations, it can in fact help you create a visual story by generating a series of related images with style, parameters and coherent moods.
To start, you just have to ask him to tell you a story and how often you want an illustration going with the action. The result will include these images accompanying the text.
For my project, I asked the AI to “generate the story of a heroic baby dragon who protected a Fairy Queen from a Diabolical Wizard in a 3D cartoon animation style. For each scene, generate an image.” I saw the start above. And, if there is a problem, you can rewrite one of the songs in history and the model will regenerate the image accordingly.
Be super precise
If you tell Gemini to make “a dog in a park”, you could get a blurred grenouet sitting in a vaguely green place. But if you say: “A golden retriever sitting on a wooden bench in Central Park in autumn, with red and orange leaves dispersed on the ground” – you get exactly what you imagine.
AI models thrive on details. The more you provide, the better your image will be. So, for the above image, instead of simply asking for a city with a futuristic aspect, I asked “a retro-futuristic urban landscape at sunset, with brilliant and blue neon signs, flying cars in the sky, and people walking in retro-filure style outfits.” Seven seconds later, the result arrived.
Convers
One of my favorite things about the new Gemini Flash is that you can do the conversation with it without losing a large part of the speed. This means that you don’t have everything in one go. After generating an image, you can literally chat with AI to make changes. Do you want to change the colors? Add a character? Mood lighting? Just ask.
In the image defined above, I started by asking “a comfortable reading corner with a fireplace, shelves filled with novels and a large comfortable armchair”. I then refined him by asking him to “get it dark with soft and hot lighting”, then followed by asking him to “add a sleeping cat on the chair”, and I finished asking it “to give the room a Victorian Victorian aesthetic”. The final result on the left is almost exactly like what I imagined and makes Gemini feel like an art assistant, capable of adapting to what I want without starting again each time.
Gemini Flash corresponds to Chatgpt
Google has boasted that Gemini is full of real knowledge, which means that you can get a historical precision, realistic cultural details and true images in life if you ask. Of course, this requires being specific. For example, if you invite him to “a Viking Warrior”, you could get something more like a Game of Thrones character. But if you say: “A historically precise Viking Warrior of the 9th century, wearing a detailed cushion armor, a round wooden shield and a traditional Nordic helmet” – you will get something much more precise.
As a test, I asked the AI to do “an ancient Mayan city at sunrise, with imposing stone pyramids, an environment of lush jungle and people dressed in traditional Mayan clothes”. It is not perfect, but it looks much more like the real thing than to previous versions, which would sometimes return with almost an Egyptian pyramid.
Write down
Most IA image models have long fought with the rendering of the text, transforming the words into illegible scribbles. Even the best models today that can do it by taking a little to do it and do things correctly can take some trials. But, Gemini Flash is incredibly good to integrate the text into the images quickly and readably. Being very specific can however help you.
This is how I generated the image above by asking the AI ”to make a vintage style travel poster which says” visit London “in a daring retro typography, with a stylized illustration of the city.”
