Midjourney V6 cracks in-image text and promises to bring more photorealism to its generations, leading the pack of the major AI models as one of the best text-to-image apps in the space.
Nine months is a HUGE amount of time in AI land... and Midjourney v6 looks like it's leaps ahead of even Midjourney v5, which was itself a huge technical feat!
@chrismessina what do you think about breaking outside of discord?
Overall its still my favorite company / the one I use on a daily basis for fun (http://instagram.com/goated.ones) and for ideation for work stuff. They desperately need to break out from Discord though, the UX limits what can be done by a large amount, and the network effects of the community are outweighed by the gains in proper UX and a controlled experience. Product marketing on the homepage page overall really needs work but they are doing so well it probably doesnt seem like something that needs fixing. V6 gave me some better results, just hoping they include the style tuner in the full version as v5 with the tuner is actually better for me than V6 without in the average case. The text stuff is pretty good, not quite as good as Ideogram.ai but solid. More thoughts: https://youtu.be/2Pnbc_-Hfzw
@chrismessina just tried this, the level of details in the pictures are insane! And it doesn't require any filters / parameters to get great results! Not surprised it took them 9 months! :)
Midjourney V6 seems like a groundbreaking advancement in the text-to-image field! I'm genuinely interested in learning about the techniques and algorithms used to crack in-image text and enhance photorealism.
It would be fantastic if you could share some examples of the output it generates.
Kudos to you for pushing the boundaries of AI models!
Thanks for sharing. This looks great. Being able to generate text correctly is a big deal.
Midjourney used to generate "shiny" images of people. It appears that Midjourney v6 has improved this aspect too.