sorry there was a google cloud outage that took our services down, everyone was working hard at 3-4 am to diagnose the issue and we got it back up as soon as possible
@midjourney
-
New Quality Modes for AI Image Generation Optimization
By
–
If you want to disable optimizations you can type –q 2 (for now). We expect the new default to have better hands so you'll likely prefer it. We also are testing a –q 4 image mode which might have improved coherence and details. More updates to –q 2 and –q 4 coming soon.
-
V7 Speed-ups and Cost Reductions Launch Today
By
–
We're deploying V7 speed-ups today along with a new default "fast mode". Standard V7 jobs are now ~2x cheaper and turbo jobs are 30% faster. Omni-reference are the same as before as we haven't optimized them yet. Have a nice weekend!
-
Niji images integration in oref v7 AI tool
By
–
probably not! but try putting a niji image into oref in v7 and it might work very well!
-
AI Model Dataset Expansion for Image Generation Tasks
By
–
We got slightly better results when doing something simple like "a perfume bottle in a forest" with a square aspect ratio, but actually it looks like our dataset doesn't have much in the way of data like this right now. We'll send it to our researchers as a good example for
-
Compute Resource Allocation for AI Relaxathon Events
By
–
We only added that for the relaxathon when we had lots of spare compute. Rn we don't have as much spare compute but will consider it in the coming months again for sure.
-
Favorite AI creations and prompt engineering techniques
By
–
congrats! what's your favorite things to make? or favorite short test prompt?
-
Combining Images in Prompts Boosts AI Model Performance
By
–
You might find if you have two people / things in one image you can prompt them both into the other. Some people are pasting/appending/concatonating two images side by side and also having luck with that.