Picture-Perfect AI: Unlocking the Key to Exceptional Image Generation: Part one!
on
Get link
Facebook
Twitter
Pinterest
Email
Other Apps
Picture-Perfect AI: Unlocking the Key to Exceptional Image Generation: Part one!
Picture-Perfect AI: Unlocking the Key to Exceptional Image Generation: Part one
Hi! I’m going to give you some tips on how to craft successful AI image prompts with Midjourney. First, my most important tip for creating brilliant images is less is always more. It works best to be clear and concise in your prompt crafting. The AI can only comprehend so much. It can understand a tremendous amount of language, concepts, and descriptions, but it’s not very good at abstract ideas.
For example this prompt:Fully face forward position. Capture a half-body shot of a breathtakingly handsome model in a studio, looking into the camera with a deep, penetrating gaze that commands attention. He sports a very cool crew cut. The man is wearing a black t-shirt and is posed against a neutral grey background. The lighting is soft and even, highlighting the shape and texture of the haircut. The background is a plain, gradient gray, ensuring that the focus remains on him. The image conveys a sense of simple elegance and is likely a professional photograph given the quality and style. — ar 2:3 — style raw
That’s a ton of text and for an AI image generator. At the most an AI text to prompt image generator like Midjourney can focus well on 5 concrete concepts, the rest it will ignore. So, decide what five key elements you want in your image and keep it to those or less. Here’s the AI generated image for a simplified version of that prompt:
Prompt used: facing forward, half-body shot of a handsome model in a studio, looking into the camera, intense gaze, crew cut, wearing a black t-shirt, neutral grey background, soft lighting — chaos 10 — ar 3:4
So, if we break it down we have about about 4 concepts in this one prompt and Midjourney gives us exactly what we wanted it to do for our AI image prompt.
1. Composition format: facing forward, half-body shot of a handsome model in a studio
2. What the person/character in the image is doing: looking into the camera, intense gaze
3. Appearance: crew cut, wearing a black t-shirt
4. The background and lighting effect: neutral grey background, soft lighting
So it’s important to note that the AI image generator may become less focused and coherent with each added element. While there are still amazing results to be achieved, we must proceed with caution in order to maintain a sense of clarity and purpose. Therefore, we should carefully consider the potential effects of adding complexity to our AI image prompt.
As someone who has studied art and taken various courses, adhering to a fixed number of concepts and elements feels second nature. Traditional art mediums have limitations, requiring artists to make decisions based on factors like time, materials, costs, subject matter, and skill level. However, some individuals exploring AI-generated image-to-text prompts believe that more context, text, and details will improve the process. In reality, the opposite is often true.
Prompt for above image used simple concepts to get beautiful results: Goldilocks as a modern bi-racial woman, fashion photography
Creating a painting requires a careful balance of colors and brushstrokes; too much and the image becomes muddled, with paint dripping and the canvas warping. Sculpting from stone is a similar process, where each small chisel stroke contributes to the final creation. You can’t expect to use a jackhammer on a rock and end up with Michelangelo’s David.
When using the AI text to prompt image generator, be mindful of copying and pasting entire paragraphs or sections from your book or story. The tool is capable of generating a lot, but it cannot interpret emotions or understand subtle nuances in your writing style. Descriptions, dialogue, setting details, and character emotions may not translate well when inputted into the AI image generator as a whole.
I have often observed individuals in the newbie groups on Midjourney using large portions of their novel as their AI image prompt in hopes of getting a specific character or landscape image. However, this is not an effective method. The AI is not able to comprehend 500-word essays from fantasy books. It is important to be concise and use fewer words, especially with the new V6 release of Midjourney. In previous versions, more effort was necessary to achieve desired results.
Another thing that will make you a better AI image prompt creator is being curious! Look at many photographs and artworks. It’s helpful to have some understanding of art mediums, photography styles, and lighting effects to help you achieve an excellent result. If you want a gorgeous portrait, do some research, look at popular images, and determine what sort of qualities you want. Do you want dramatic, backlit, close crop, cinematic, zoomed out, etc? Find a detailed guide on crafting AI image prompts here for medium users and for non-subscribers here.
Understanding depth of field and perspective can greatly enhance your AI image prompt. Decide if you want to highlight objects in the foreground or background, and communicate this to the AI for different effects such as blurriness or sharpness. Experiment with different ratios, such as the default square, to see how they influence the overall composition of your image.
Though now with V6 you can zoom out, as well as pan in different directions. So sticking with the default square means you don’t necessarily have to pick a particular ratio. But if you’re looking to do sweeping epic images, then I recommend ar 16:9. If you’re looking to do a sky scraper, or height oriented concept, you could try ar 9:16. For portrait style ar 2:3 lends itself well to that type of image. With the new MJ tools you can revert images back into squares, so plenty of customization is available.
So you don’t need to understand the ratio as much as you used to before, but it’s still helpful to achieve a particular result in your first prompt generation. For example, if you want a character for a book or a game, try starting with the ar 2:3.
Okay, that’s a wrap on part one! In part two, I’ll share some prompts so you can try them out! Happy AI generating!
Comments
Post a Comment