The painful issue with all generative AIs is resolution. Despite significant progress in generating larger images last year, artificial intelligences (such as Midjourney, Dall-e 3, and others) still primarily generate small images suitable for web use.
However, by following certain procedures, it is possible to obtain results from artificial intelligence that can be easily printed in high-quality photographic format on a large scale. I have experimented extensively with enlargement techniques, and the following is my preferred process, which usually yields the best results:
1. Generate the desired image using artificial intelligence (At this stage, it is recommended to consider that the image will be enlarged. The fewer details in the image, the better the result. This is because larger areas without intricate patterns will enlarge more effectively. Therefore, I often use a shallow depth of field - for example, I use lens f1.2, telephoto lens, etc., in the prompt. If you are working with Midjourney, use version 6.0.

2. Perform the first image enlargement directly using the upscale command in Midjourney.

3. Save the resulting enlarged image to your computer, open it in Adobe Photoshop, and retouch anything that the AI did incorrectly or that does not meet your expectations. Then save the image in TIFF format. In this step, I often use Adobe's generative AI to add additional surroundings, further enlarge the image, and adjust it to the desired aspect ratio.

4. Using Adobe Bridge, open the created TIFF file with Camera RAW. Choose the Enhance - Super Resolution function. Camera RAW will further enlarge the photo 4x and save it in DNG format.

5. Again, open this new file in Adobe Photoshop. In this step, fine-tune the details. Some photos may need sharpening, while others may require slight blurring. Sometimes adding noise can unify the entire image.

6. After final adjustments in Photoshop, I often slightly reduce the size of the photo, which helps smooth out minor imperfections caused by sharpening and noise addition.
Download final 100Mpix AI Photography - HERE

To demonstrate the effectiveness of this process compared to others, I have provided a comparison below. The first image was generated in Midjourney, the original was enlarged using the bigjpg application, and the final enlargement, additional surroundings, and detail refinement were done in Adobe Photoshop.
The second variation showcases my tested process described above. As seen in the zoomed-in excerpt from the photo at maximum resolution, the result obtained using my enlargement method contains significantly more details.
Back to Top