Midjourney v6 Tips and Tricks. Text. Chaos. Weird. Stylize. Part 2

Рет қаралды 13,056

Күн бұрын

Second deep-dive into Midjourney's latest version 6 and it's features. How to get text to appear in your images and tips on what currently works and what does not. What do parameters chaos, weird and stylize do in the newest version?
Part 1 of this series here: • Midjourney v6 Tips and...
Music generated with Mubert: mubert.com/render/pricing?via...
All generated images created by Thaeyne with Midjourney: www.midjourney.com
Voiceover generated with: elevenlabs.io/
Talking images with: www.d-id.com/
Moving images with: runwayml.com/
Also moving images with: kaiber.ai/
0:00 Intro
0:23 Text in v6
2:49 Chaos
4:30 Weird
6:33 Stylize

Пікірлер: 23

@boerboelmummy5075 4 ай бұрын

Thank you, and a very happy New Year to you.

@nic-ori 4 ай бұрын

Thanks! Useful information.👍

@drfranke 4 ай бұрын

Happy New Year, Thaeyne! All the best in 2024!!!

@cliftonux 4 ай бұрын

Weird and Chaos are useful. Particularly in combination with Style. Adjusting the number values in remix mode always gives new options. It’s best to leave all three in with a variety of values if you want surprises. Sometimes the combined results of one higher value and two lower values gives fantastic new directions to branch off into new ideas.

@ohshi013 4 ай бұрын

Everyone tends to forget that the current V6 is only a progress alpha version (which was more likely pushed out to only satisfy the impatient fanbase and keep the current subscribers from cancelling their subscriptions), of course it will have issues and bugs because it's not finished. Everyone should not treat alphas, betas, etc. as completed versions. Yes, there are issues, distortion and missing previous features but again, this is an alpha which should have only been released to 10K plus image subscribers first. Additionally, the chaos and weird features only work well in the raw style usually up until a '15 - 20' value at most, thereafter the prompts original intent gets lost.

@ohshi013 4 ай бұрын

Check me out at @ohshi13, I have been able to create some good style raw images with '1 - 20' chaos and weird values in V6.

@mtntopviking 4 ай бұрын

Well said. People are getting too worked up over what isn't included - YET. It will get there. I like it is so far and am looking forward to more and better updates. I switch back and forth between the website and discord depending on what I am doing and looking for. The features I miss most in version 6 is, "inpainting," and, "zoom." If what I am trying to create will more than likely need them I use discord and version 5.2.

@pho6149 3 ай бұрын

hi! this video was very helpful to me, thank you. can you please explain why do you have --seed 777 in all requests? what does this give?

@christopherd.winnan8701 4 ай бұрын

Great video as always. A few suggestions. For the talking heads, have you tried cultures where heads are held more steady, such as Thai Hill Tribes or Kisaeng? I like the use of Mythpunk, and this was the first time i have seen that combination. Any chance you can do an extended x-punk styles video? I have seen that celestial and cosmic work well, as do combos such as oriental and steampunk. I was wondering about platinum and shanzhai? What new punk styles would you suggest? Finally, keep a hold of those comparison images. They are quite difficult to analyze at the moment, but once we have a few more capable multi modal models, you will be able to feed them in and learn all kinds of amazing new insights!

@thaeyne 4 ай бұрын

I am in the process of testing out the styles and probably will make some videos soon about some of my favorite outcomes of those tests. I also liked the mythpunk a lot this time around. v6 really has a new and different spin on some of the styles that I have already previously tested with v4 and v5.

@RassamFanan 4 күн бұрын

-- style raw works with text better especially if the text is non-english dictionary defined word

@egauci1 4 ай бұрын

V6 still needs a lot of work. We need things like pan and zoom back, and it's currently too slow. Still it's a huge improvement over v 5.2, especially in its ability to follow prompts with multiple elements. For example: "spooky mushrooms and jack-o-lantenrs on Halloween night. Bats in silhouette fly in the background under a full moon. Children play with their pumpkin treat baskets", while making beautiful images in both v 5.2 and v6.0, produced absolutely zero mushrooms in v 5.2. This was an area where Dale-e 3 was much better, but Midjourney has caught up.

@EverSoInnocent 4 ай бұрын

Would love if you could experiment with grayscale and black and white coloring pages. It's so difficult to get a decent coloring page. If you try doing black and white only with bold outlines, it still tends to give you too much shading (specifically on "adult coloring pages." Same with grayscale. It's just TOO much shading, no matter what parameters you input. I tried experimenting with TUNE, but it's been so all over the place it hasn't been worth the cost of fast hours. Starting to feel like MJ is falling way behind other programs, but I'm reluctant to start all over with learning a new AI program. I feel like MJ's focus on barely useful features like "weird" is really setting them back.

@Sanchan 4 ай бұрын

When I used an picture with text on Niji, It doesnt show up correct when I used quotations in the prompt. I assume text got better only on V6?

@thaeyne 4 ай бұрын

Yes, Niji is not running on v6 yet and text works strictly on v6 at the moment.

@spacekitt.n 4 ай бұрын

the weird talking characters in these videos are creepy and unnecessary

@Thoumint 4 ай бұрын

Lol

@christopherd.winnan8701 4 ай бұрын

I agree that the examples featured in this video are not up to the usual standards and do indeed look a little jarring. Might I suggest adding the following adjectives to the prompt to encourage head stability - haughty, aloof, refined, Swiss finishing school. This should introduce more steadiness, but hopefully without hallucinating any actual book balancing.

@thaeyne 4 ай бұрын

I wanted to leave them in, because without them this video would have been just a slideshow of images. But I didn't also want to do animations with Kaiber or Runway this time either, because those transform the images so much that they are no longer v6, but stable diffusion instead.

@thaeyne 4 ай бұрын

Noted. I didn't specifically prompt for any images that would be more better for the heads. Instead I wanted to try out as much styles as possible from the ones I have tested out with older models and then select some images that could be possibly suitable for the talking heads. But the selection came out small as v6 doesn't do as many heads anymore and they are more often than not really close up or not centered or have other things wrong with them that don't really make for good material. I will do my selections more carefully in the future and maybe do less of them. And still as I just mentioned in a previous comment, I didn't want this video to just be a slideshow of images and I also really like experimenting with the movement for the images as well as much as possible.

@christopherd.winnan8701 4 ай бұрын

@@thaeyne Please keep them coming. They will get better over time.