I couldn't resist giving the new v5 version of MJ a test drive. Here are my best results.
The chocolate mess shots were entirely text-based prompts - no image prompts were used for those. While I was able to get some very interesting and pleasing environmental looks with that approach, it was extremely difficult to get any substantial facial coverage, no matter what I tried. v5 is in alpha "pro" mode at the moment, though, so maybe things will improve in time.
Those are great, made me play around a bit with V5 of midjourney too ... a big improvment, though it is still hard to get consistent results. I only managed to get good coverage with supporting image prompts
MidwestMucker said: I couldn't resist giving the new v5 version of MJ a test drive. Here are my best results.
The chocolate mess shots were entirely text-based prompts - no image prompts were used for those. While I was able to get some very interesting and pleasing environmental looks with that approach, it was extremely difficult to get any substantial facial coverage, no matter what I tried. v5 is in alpha "pro" mode at the moment, though, so maybe things will improve in time.
WAM Photography, that is some amazing work! Two questions:
1) How are you doing this? Mind writing a quick how-to guide for dummies? Especially in terms of getting an AI generator to understand that you want the woman to look both smart AND messy? That's the part I'm struggling with.
2) I'm massively into women wearing ties and looking smart and formal while getting gunged or pied, so seeing your work is amazing, but I've also been a fan of your real-life stuff as well. Ever thought about doing a gunge / pie video with a woman who's wearing a perfectly knotted tie as well as a smart business suit?
FormalGunge said: WAM Photography, that is some amazing work! Two questions:
1) How are you doing this? Mind writing a quick how-to guide for dummies? Especially in terms of getting an AI generator to understand that you want the woman to look both smart AND messy? That's the part I'm struggling with.
2) I'm massively into women wearing ties and looking smart and formal while getting gunged or pied, so seeing your work is amazing, but I've also been a fan of your real-life stuff as well. Ever thought about doing a gunge / pie video with a woman who's wearing a perfectly knotted tie as well as a smart business suit?
Thanks!
I also want to know. These scenes were sensational. Could ever be sold as real WAM and nobody would see the diference.
Okay, my first post in 18 months or something like that. I've been tinkering with MidJourney too. This is the future. Get ready for it. Threw mud all over Marilyn Monroe's face. I'm going to put her in a Tarzan movie before this is over.
DuncanEdwards said: Okay, my first post in 18 months or something like that. I've been tinkering with MidJourney too. This is the future. Get ready for it. Threw mud all over Marilyn Monroe's face. I'm going to put her in a Tarzan movie before this is over.
Question for those experimenting with MJ v5, are you paying for the basic sub? I want to try it out but I'm concerned only the top sub allows a private mode which isn't posting in the public discord channels. Or do I misunderstand how it works, I find discord a bit alien coming from traditional forums.
The T&Cs say no fetish images so aren't you getting reported?
SmushingTin said: Question for those experimenting with MJ v5, are you paying for the basic sub? I want to try it out but I'm concerned only the top sub allows a private mode which isn't posting in the public discord channels. Or do I misunderstand how it works, I find discord a bit alien coming from traditional forums.
The T&Cs say no fetish images so aren't you getting reported?
I have a subscription for now. I don't really care what anyone else sees while they are there. MidJourney won't let you use anything in word or image that can't be in a 30 year old Disney movie. I'm sure they will open things up a bit as it matures but right now it would probably reject the word "fetish" instantly. They will give you 25 trial runs images if you want to play with it. Like so many things you can begin with just typing a few words but it's a little more effort to refine images. I've been amazed so far at what it will do I'm just getting started. At another board I'm part of, someone discussed a news story of being paid to chase seagulls off the mudflats near a beach. Took me about a minute to create this -
Could you detail a bit about how you got the promps made? Did you supply midniurney some examples of your previous work for a base ?
FormalGunge said: WAM Photography, that is some amazing work! Two questions:
1) How are you doing this? Mind writing a quick how-to guide for dummies? Especially in terms of getting an AI generator to understand that you want the woman to look both smart AND messy? That's the part I'm struggling with.
2) I'm massively into women wearing ties and looking smart and formal while getting gunged or pied, so seeing your work is amazing, but I've also been a fan of your real-life stuff as well. Ever thought about doing a gunge / pie video with a woman who's wearing a perfectly knotted tie as well as a smart business suit?
Thanks!
rt2018 said:
FormalGunge said: WAM Photography, that is some amazing work! Two questions:
1) How are you doing this? Mind writing a quick how-to guide for dummies? Especially in terms of getting an AI generator to understand that you want the woman to look both smart AND messy? That's the part I'm struggling with.
2) I'm massively into women wearing ties and looking smart and formal while getting gunged or pied, so seeing your work is amazing, but I've also been a fan of your real-life stuff as well. Ever thought about doing a gunge / pie video with a woman who's wearing a perfectly knotted tie as well as a smart business suit?
Thanks!
I also want to know. These scenes were sensational. Could ever be sold as real WAM and nobody would see the diference.
Sorry for my late reply guys, I missed your questions since I'm rarely on these days.
Most of the time I combine an image prompt with a text prompt. I experiemented with shots of my models and oftentimes use two images. Usually I use one clean image (first one) that tells the AI how I'd like my subject to look like and then a messy one that matches my text promp as closely as possible and where the look is not too far off the clean image. It does not necessarily need to be the same woman. If you use a celebrity in the text prompt the system will have a better image base to draw from.
So I experiemted quite a bit with Teri Hatcher cause she looked sexy as hell as Lois Lane in the old Superman series. And there's quite some images from that time the AI can draw from. Midjourney got better with every version in making the images look more natural.
I added a comparison from v4 to v5 you can see the face looks much more natural in v5. I tried to recreate exactly the same image (same source, same prompt) in both versions. I also added some of my recent tries with v5.1
I added a few more tries ... I focused on chocolate this time and my AI models enjoyed the choclate coating ...
I worked with wieghted image prompts this time, but decided to generate the image prompts also in midjourney. So the models as well as the chocolate mess were created iteratively in midjourney.