🖌️ AI Art Tips Thursday #18: Let's bash some photos together!

💡 AI Art Tips for the Day

By popular demand, I am going to explore how you can do photo bashing within midjourney.

The concept of photo bashing is simple. Basically, you add an element on top of an existing image and try to create a final image that would smoothly include the element on the image. Here is ChatGPT’s explanation of this: Photo bashing is a technique in digital art where artists use existing photos, editing and merging them, to create detailed and realistic elements in their artworks quickly.

In order to do photo bashing, you would need a photo editing tool. In my case, I have used Figma for this.

For the examples today, I am going to try both bashing with an object and a human to see how they look.

Before looking at the results directly. The following is an example of how the bashed image looks like👇️ 

I know, it does look a bit weird to just put the death star there. But that is you bashed the photos to be ready for midjourney.

Similar to how you prompt with an image link, you would upload this bashed image and get the link to be put in your prompt.

So the images below are my try to put Death Star into the sky in this image with the house and a tree.

The results are all looking good but they are not fully taking into all the details of the Death Star. Instead, the generated images have included a moon in the sky, which is pretty close. So I guess the generated images from the bashed image won’t take all the details for the bashed object. Based on the examples, my understanding is that the AI is trying to understand the bashed object and would give its best guess and add that to the image.

Cool, so now let’s try with a human being. In this case, I just grabbed a person’s face from the internet using this interesting service: https://thispersondoesnotexist.com. And the following are the results 👇️ 

I personally think they are pretty good. Also, my wife thought some of them did capture the details of the person’s face in the generated images as well. 👍️ 

So I guess faces really work well with the photo-bashing technique. It is something you can consider if you want to create a consistent character across different images.

In conclusion, I believe photo bashing is a valuable technique you should include in your midjourney promoting repertoire. A good use case of it now is to get a face reflected in the generated images. If there is another good use case you can come up with, give me a shout!

🖼️ Image of the day

Although it didn’t get the death star in the sky, I am just amazed at how much detail is included in that moon and the other objects in the following picture. Truly mesmerising 🤩