TL;DR
- A Reddit user has shared a set of compelling bad food photos, but all of them were generated by AI.
- The Redditor has shared the prompt that they used to generate these photos in Midjourney, and it’s easy to recreate the same effects for some impressive iPhone-like AI food photos.
AI image generators are one of the most prominent ways we can use AI in our lives. Feed in a text prompt to get an image that fulfills your needs. But there is quite some level of skill involved in getting a good AI image out of an AI image generator, and you’ll need a fair few attempts to craft a great prompt that works out for you. If you ever wanted to create AI images that look shockingly deceptive, here’s a prompt that can get you some very convincing iPhone-like results with Midjourney.
Redditor MagicJourknees posted this set of images on Reddit with the caption, “My GF puts raisins on EVERYTHING! Is this normal!?”
The images caught many people off-guard, who were perplexed by the person’s sheer ability to enjoy every meal with raisins. However, many noticed that the images were posted to the Midjourney subreddit, meaning that none of these images are real.
Yes, all of these images are AI-generated using Midjourney, but the style of capture and the output make them look shockingly convincing. You could show these photos to anyone without context, and they will have a hard time spotting these as AI-generated images.
Thankfully, Redditor MagicJourknees was kind enough to share the prompt they used to create these images. Here is what they used in Midjourney:
Code
iPhone photo of (food name) with many raisins on top. At a (type of) restaurant (or other location). --ar 3:4 --style raw --s 75
PLUS --sref (image URL of some bad food photos you find on Yelp or elsewhere)The initial bits of the prompt are pretty self-explanatory, and you’ll need to input the variables present within the () brackets. The –ar parameter defines the aspect ratio that Midjourney uses (3:4 in this prompt), while the –style raw parameter and argument give out a less “beautified” and more “accurate” image. The –s parameter is the stylize parameter, taking the value 75 in a range of 0-1000 with the default being 100, so the images generated take less artistic freedom and stick closer to the prompt. Finally, the –sref parameter lets you use an existing image as a visual style reference, and inputting a direct image URL gives you a solid reference point for your generated image.
Try out the prompt above to create some very convincing bad food photos with Midjourney, and let us know your experience in the comments below!
