It’s not just about MidJourney’s SREF codes. Explore a universe of styles and visual models across platforms Do More Than See, Say, Save, Show, Share, Scream...

Joined May 2025
Sref Code retweeted
V8.1 is live! Our iconic aesthetics are back w native 2K HD rendering - 3x faster and 3x cheaper vs V8. Full quality V8.1 1K mode is faster than V7 draft mode. Image prompts are back. New "Describe" is live - and you'll love our new moodboards & srefs. More soon <3
143
181
91
1,402
435,994
Sref Code retweeted
Today we're starting to test an early version of our V8 model with our community. It's much better at following prompts, 5x faster, has native 2K modes, improved text rendering and the best personalization, sref, and moodboard performance ever. Have fun!
184
183
137
2,298
2,952,852
Sref Code retweeted
A Style is Worth One Code Unlocking Code-to-Style Image Generation with Discrete Style Space
4
18
3
102
12,679
🚨 MIDJOURNEY UPDATE : SREF EXPLORER 🚨 Midjourney launched an early version of their style explorer today. It's like the video and image explore pages but now it's styles!, Each thumbnail is a set of prompts rendered with a different SREF code (which sample randomly from the continuous latent space of all styles). You can save styles you like by pressing 'like'. You can hover over styles and press 'Try Style" to render whatever is in your prompt bar (so you can just go through a ton of styles and render "dragon" for all of them, etc), You can also 'fuzzy search' through prerendered styles with the upper right search box. For example if you type "photographic" or "anime" it will try to find the srefs that best correspond to those words, There's more updates coming for this feature and they'll try to add more srefs continuously. Link in first comment . I am not a fan usually of srefs I never use them almost but this feature is really good , helpful, the fuzzy research is great . Hope they will change the way to save them later
9
16
1
154
11,196
Sref Code retweeted
THE EXCLUSIVE JSON PROMPT PACK IS HERE! 🎉 I just put together a special collection of 35+ JSON prompts you can use to retexture images with ChatGPT-4o. Think cool materials like transparent plastic, fluffy fur, shiny gel, voxel blocks and a bunch more.. If you've been trying to get cleaner, more creative, more consistent image outputs, this is for you! Subscribe to me and get 30% OFF the full guide! Here’s the link 👇
15
23
2
400
226,237
We've released the new V7 version of Style References and moodboards. It's smarter and more precise than ever. There's also a new --sref random with significantly more variation to explore. If you want to fall back to the old system or sref codes please use --sv 4. Have fun!
84
119
58
1,001
174,243
Cyberflora Dreamcore MJ Sref share is back! I currently have a bit more flexibility and have received some support, so I've decided to restart this project. I'll leave the sref code on the comment👇
11
6
3
93
5,344
Sref Code retweeted
Replying to @_akhaliq
@_akhaliq 🚀 Open-source breakthrough! We recreated GPT-4o-level stylization consistency using only 2,600 pairs + 500 GPU hours! Introducing OmniConsistency: ⚡ Super strong style + content consistency Demo 👉 huggingface.co/spaces/yiren9… Code 👉 github.com/showlab/OmniConsi…
1
3
84
Sref Code retweeted
Tried training the isometric 3d style. This is after ~40 runs. The prompt to image pipeline is chaotic even at the lowest variance settings. -- Realized there's a pattern to image generation across all major ai providers. - Train the style first - Then add a reference object or prompt - Refine output till you are happy (hopefully) Working back from an expected outcome is a mindset shift to how creatives usually work. --
I got a lot of questions + DMs about 3D icons like this. Turns out, generating either ultra simple or complicated renders using AI is a bit tricky. Here's what worked for me. I used the Airbnb icons as a reference. For 3D icons I generate sequentially, I re-use my generated results to keep training the chat to reference the style again and again. If I am unhappy with a texture, I ensure I point it out specifically with the desired outcome. Again, for such 3D icons, I generally format a prompt based on a syntax. Angle, texture, details, object and any other key attributes I want factor in too. I used Ikea images (1 product shot, 1 scene shot) to ensure the lighting, glow and materials were right. I name my references for example, Lamp 1 and Lamp 2 here. In the end you can generate endlessly, by just replacing object reference, texture, colour details. Hope this helps! For Airbnb style icons I did refer their iconset, but for other 3D icons it took a lot more effort to prompt simply and repeatedly, no external reference was used.
2
1
9
1,631
I've just published my next Article Link on the thread 👇 Artists mentioned in this article with their Srefs: Felix Koutchinski @Koutchinski The_lostastronaut @dlostastronaut Now my articles are on Kinomoto Mag in Medium. Thanks @KinomotoMag for the invite!
3
5
1
24
1,286