Video Summary
Here are the key points from the video:
- Crafting better prompts is essential for creating better images with generative AI tools like Microsoft Copilot. The more specific you are, the better the results will align with your expectations.
- You can enhance your prompts by including details such as colors, background information, art style, lighting, framing, and camera settings. These elements help Copilot understand exactly what you want.
- Specifying the art style is crucial. For example, you can ask for a realistic image, a painting in the style of Vincent Van Gogh, a pencil sketch, or even pixel art. Each style will produce a unique result.
- It’s possible to mix different art styles in a single image. For instance, you can combine photo realism with 3D design and a painting style in the background.
- Every detail matters when crafting prompts. Whether it’s the color of a dress or the type of lighting, specifying these details ensures that the generated image matches your vision as closely as possible.
For more information, read the transcript blog below, or watch the video above!
Video Transcript
With the first few examples done, let’s take our image generation skills to the next level and create better images. But how do we do this? The answer is we need to craft better prompts. Since generative AI tools cannot read our minds—well, at least for now—we need to tell them all the details of what we want them to create. In our case, Microsoft Copilot. The more details you add in your prompt, the better the result will align with your expectation. You will often see Microsoft talking about adding more prompt ingredients to your prompts. But what are some of those prompt ingredients that we can add? We can add things such as the colors you want in your image, specify background information, the art style you want—that is very important. You can also specify the type of lighting that you want, the framing, and even the lens and camera setting that you would have used for the image you want. Let’s deep dive into some of those.
Let’s start with art styles. It’s very important that you specify the type of image you want from Copilot as it will make a huge difference. In this example, I’m asking Copilot to create an image of a white middle-aged woman walking on the beach near the ocean at golden hour as a realistic image, and this is the result. I can also ask the same thing but as a painting. Actually, I want it more specific, as a Vincent Van Gogh painting. So you can tell Copilot not only that you want it as a painting but give it the style of a painter that you like, or specify that you want it as a watercolor painting or as an oil painting. We could also ask Copilot for the same image but in the anime style, which will give you something similar to this image. I can also ask Copilot for a pencil sketch of my image. Again, we are using the same prompt; it’s just the image style that is different. And here I was even more specific. I asked for a black and white pencil sketch. I could also ask for the same image in pop art style, which would give me something similar to the image on the slides here. And while I could go on forever as we have thousands of styles in the world, here is a last example of our same image but in pixel art style. So as you can see, the past six slides have all been asking Copilot for the same exact image but in different styles, which creates completely different images.
So what I want to remind you is that every detail matters, from the art style to the lighting you want, to the depth of field and type of lens. Really tell Copilot all the objectives that come to mind as you imagine your design, your image. This way, the result that Copilot generates is as close as possible to what you expect. Now let’s try some of those more advanced tips and tricks that we have learned. For the first example, I will use the prompt that I did in the slides: create an image of a white middle-aged woman walking on the beach near the ocean at golden hour as a realistic image. So this is the photo-realistic image option or art style that we saw in the slides. And remember when I said that most of the time you will get four options? Sometimes you will get just two, and this is what Copilot came up with. Remember, every detail you give Copilot matters. So, for example, if I want her to be wearing a black dress, I need to tell Copilot. So if in my head I was imagining that that woman would wear a white dress, I need to tell Copilot in order for Copilot to give me the image that I wanted. And there we go. Now it generated exactly what I wanted.
We mentioned art styles, and I always gave examples of a single art style, but if you want to, you can actually mix art styles. In this example, I’m asking Copilot to generate an image of a photo-realistic Asian man placing an order in the kitchen to a 3D design robot kitchen chef. The wall behind has an ocean painted in Monet style. And let’s see what it did. Here, I am not an artist, so I cannot tell you if this is proper Monet style, but you see we mixed the photo realism with a 3D design and painting all in the same picture. So you can mix art styles as well. Now, what are some use cases for using Copilot generation?
