In the beginning was the word, and the word was with prompt, and the word was prompt. Then there was an –sref code…

If you’ve been browsing social platforms lately, you might have noticed an intriguing new trend in the generative AI space: those mysterious codes! At first glance, they seem like random strings of characters, but when used in a Midjourney prompt, they produce a somewhat predictable visual style. How random are these codes, what value do they hold, and, most importantly, how do they work? Let’s take a deeper dive in.

Here are some important details about the –sref codes and their role in stylizing images. According to Midjourney’s release notes, the best way to explore the extensive range of style references they have added is by using the –sref random feature. This allows users to delve into the “latent space of visual styles.” Once you discover a style you’d like to explore further, you can use –sref url to generate more images in that style. Here are a few key points to remember about the codes, as shared on X by Alie Jules.

  • –sref codes are seeds
  • style reference seeds a.k.a. –sref code allows you to replicate the style of that seed when you use the same –sref code with various prompts.
  • can’t be combined together (possible in the future)
  • can’t be used together with style you apply from an image
  • these current sref codes won’t work in v7 (however they will work in v 6.1 which will be coming soon.)
  • you can use style weight (sw) with srefs.

Now that we’ve covered the basics, we decided to share some of the most intriguing –sref codes we’ve discovered. These were found by following some of the most notable and creative genAI adopters on X. For detailed prompts please use the resources links. Here we go!


curated by: Marco / source: X


curated by: OscarAI / source: X

TRIPPY –sref 3787560512

curated by: Alie Jules / source: X

Alie has also begun developing a comprehensive –sref codes library, which is a must-see if you want to stay ahead of the game. See here.

SENSE OF MOVEMENT –sref 2421610009

curated by: Gizem Akdag / source: X


  • –sref 4072648137 (moody/ soothing atmosphere) /
  • –sref 410263441 (colorized image)
  • –sref 3972221354 (ambiant haze & pinkish hue)
  • –sref 2691217566 (classic 35mm feel/ earthy tones)
  • –sref 1234437726 (post apocalyptic/moon atmosphere)
  • –sref 1798194133 (low-quality/amateur type of pictures)
  • –sref 1028682966 (daylight pastel colors)

curated by: Hugo Ventura / source: X


curated by: Nick St. Pierre / source: X

And a crown in the jewel from Nick himself.

“I just made my own style reference explorer for Midjourney Used AI to automatically tag and label the images so I can search them by style, color, lighting, etc and find style codes that fit that vibe.

All 1500 style references are over in my discord and you can browse them like a visual dictionary.”

Available here.

RETRO-FUTURISTIC –sref 3038856255

curated by: Dogan Ural / source: X

MINIMALIST –sref 3417747545

curated by: Tatiana Tsiguleva / source: X

And another gem here. She reviewed hundreds of SREF codes and identified 15 high-quality ones that work well with various subjects and look great in both photorealistic and minimalist styles. She compiled all of them into a beautifully designed PDF, ready for you to download here.

80’s MOVIE –sref 2158083008

curated by: MayorkingAI / source: X

Did you know?

There is a definite number of –sref codes in Midjourney. The total of 42,949,672,095 –sref codes in Midjourney makes the library quite rich, doesn’t it? Why the number? Apparently it is the largest 32-bit unsigned integer (2^32 – 1). In hexadecimal notation, this number is represented as 0xFFFFFFFF, which is the highest value that can be represented with 8 hexadecimal digits, according to Derya Unutmaz, one smart X user who shared the story behind the logic*.