ALBUM COVERS REIMAGINED USING AI
THE CONCEPT
After stumbling upon a the “100 Worst Album Covers Ever” on Rate Your Music, I reimagined these admittedly awful album covers while maintaining the overall aesthetic of the bands and artists. My intention in this project was to experiment with AI tools, namely Midjourney and DALL-E), to generate the primary assets for the album covers. I aimed to stay away from any vulgar or suggestive imagery, which, in my opinion, is one of the main reasons why these album covers have aged poorly.
THE APPROACH
I engineer these images using two prompt techniques: The first technique is the 2-4-2 approach, where you begin the prompt with the subject + setting followed by detail + detail + detail + detail and finish the prompt with mood + atmosphere. The second technique is the instructional approach, in which you structure the prompt using short sentences to set the scene. This is currently one of the most complex prompt techniques. I used a combination of these techniques to maintain tighter control over the process while keeping the prompt structured and easy to modify. This further helped me identify any flaws in the prompt, such any description the AI is ignoring. Of course, there’s no right or wrong way to prompt but rearranging the prompt can greatly affect the end result. Another trusted technique is using affirmative language when instructing the AI. It is best to avoid negative instructions, e.g. “Do not include any dark imagery,” because this is nonspecific and may introduce ambiguity. I further edited the images in Adobe Photoshop to put the finishing touches on the album covers. Finally, I used Adobe Illustrator to create the textual treatments.
