The Midjourney V5 model is the newest and most advanced model, featuring very high Coherency, excelling at interpreting natural language prompts at higher resolutions, and supports advanced features like repeating patterns with --tile.
The image quality performance of v5 is mind blowing. I can't believe how good these AI image generators are getting as fast as they are.
Check out this v5 video... Practically Pixar level:
https://twitter.com/nickfloats/s...
It's getting harder and harder to decipher what's real and what isn't. Midjourney V5 seems to have doubled down on everything it's learned, for me, the most stand out aspect is hands, hands are finally fixed and I'm not sure how confident I'm going to be telling AI art vs Human Art anymore.
@alvaroybanez good question! Not necessarily, I’m not super alarmist about it but I do believe people need to be a little more skeptical of photos we see online now, people have agendas and will try pass some off as real
@aaronoleary gotcha. I'm on the side of liking something regardless of the source. I don't think I will be influenced by the agendas you mention, but I can see how someone would like to win the National Geographic contest with it. :(
@aaronoleary For simple idea prompts hands are fixed, not 100% though for more complex scenario prompts. I love MJ5, but I do miss being able to fix things by upscaling, a few eyes here and there are still coming out funny - OAI say they're bringing back a more arty Midjourney with the choice of toggling between that and the new more accurate rendering. What a great time to be alive.
you know how explicit music has a parental advisory council badge and Getty images has that huge watermark? I'll need 1 for AI generated images.... they are getting too real 😳.
Is it still $30? I avoided this since the beginning due to transparency issues back when they used VQGAN+CLIP and Disco Diffusion without saying so, before SD was released and they moved over to that, still obfuscating their backend as if it was their own.
Unbelievable mind-blowing image generation - Midjourney has revolutionized the whole creative process for me. Coming from other tools, the AI-generated art is stunning and high quality even with simple prompts, but it's more than that because of the training and customization.
There's levels and modes for stylizing/quality, you can learn by doing and/or by seeing what others have created. The discord is one of the best communities I've seen for learning/collaboration.
The docs are great too - there's more parameters, but it's a balanced view of how a tool should work - enough style and aesthetics/quality with power to change all the parameters and control. V5 is less opinionated than other versions, but the prompts can be used more like a chat, and less keyword packing/syntax. I use v3, v4, and v5 for different purposes so that's an option too.
One of the biggest updates is to the hands 🖐️ for all the AI haters (they are funny tbh) but the hands are getting better and more realistic.
It keeps getting better the more I use it and discover. Just scratching the surface for the last few days, it gets better all the time, big shout-out to the team that is working on this. 🌟🌟🌟🌟🌟
It's such a great improvement over v4 to me, especially for realism! The one thing I wish Midjourney had was a proper web interface, even a simple one built upon Gradio would be a step ahead of Discord I feel like.