I trained a stable diffusion lora model on mud photography which I have purchased. It is like 90% Mud Puddle Visuals (plus Muddy Pinay's / Tropical Shots Media). There are a few images from other produces. I also included some random images from my collection from photos I grabbed from forum previews since like a decade ago.
I labeled this nudity because the forum posts here so far have been non-nude. I don't know if this group was trying avoid that.
Captions:
Co-Worker - who wants you to join her in the empty office break room to wildly play with mud with you before you two get caught and fired. Potter - who prefers to be naked because she "accidentally" gets a lot of pottery clay on her body while she works. Traveler - who cannot take the jungle heat, sweating, and needs to stripe and dip in a mud pit to cool off. Gardener - who loves flowers and wishes she could grow them from her muddy body. Friend - who is revealing a tent that she prepared mud in and she wants you to join her tonight. Scientist - who tested her secret massage clay mud formula and is surprised in how good it feels on her skin. Singer - who wanted to give a unique performance by doing it covered in mud and covers her privates with her hand.
I feel like I could write a story for each of these images. Any of these could be tweaked in many ways or turned into a photo set.
I didn't ask permission from the producers to use their photography. This is just a hobby for me to have fun with. The images I generate could very closely resemble Mud Puddle Visual images. That is not the intent. These are some examples.
I am going to include more content in a v2 model. I recently just subscribed to Muddygirlies.I really want to get 3000px images and pre-2019 content from Muddygirlies, but that was taken down with the March 2023 site redesign. I'm going to keep subscribed until I get those images as well as the newest stuff.
If anyone reading this knows of any other producers content who offers photography and not screen captures, I am interested. If the resolution and frame rate of videos is high enough, it is possible to use some images from screen capture. I cannot use anything blurry, and videos of motion are not very good to try to get screen captures of. Sometimes I need specific concepts which there are no photos of, so it may be necessary to use screen captures. For example, I don't see a lot of pictures of red clay. There are a few scenes of MessyGirl and MudMadPhils that I might try to include more of to get a good red clay.
I have no intention on sharing the lora model. Sorry. I cannot control what others do with it. I have no intention of trying to monetize from it. If it is shared, someone might do something like open a store and try to sell photo sets or even videos. They can do it deceptively. This is a bit extreme, but I think it is the fair thing to do since I did not get permission from every producer. If any producer does not want me to use images in training, I will simply stop posting images.
I want to go a little rant here.
This is not meant to replace the photographs of real women. I am doing this purely out of creativity. Producers can use any of the creative concepts which I create.
This is not this meant to deceive anyone even though the photos may look realistic. I will only share photos in this group or in my photo album, not in the main messy forum.
This is is only meant to show pictures of adults. All of the images I used are from verifiable sources. All of the prompts I use contain positive keywords for adult women. I do not use the terms "girl", "18 year old", "teen", or anything similar in my prompt. I used negative prompts to remove non-adult imagery as well. I have no control over the age of the woman of the photos I generate. It could be someone who resembles a 20 year old or a 30 year old. All of the images I generate have invisible watermarks.
This is not meant to dehumanize women and turn them into objects which we can generate. All of the images which I generate, if they were real, would be respected, would have free will, equality, and would have chosen to be photographed in the manner that they are. I also have tremendous respect for the real women here on this site.
This is not meant to reflect on an ideal size or race. I tend to generate images of women in a particular size and race that I'm attracted to. Women of all sizes and races are beautiful.
These images show some of the capabilities. Hopefully the legends are clear. The numbers are weights applied to prompt terms which can be used to control the coverage amount of clay mud on her body. I also included an image showing that it is possible to mix substances that I did not train on and where already usable in the base model.
Some notes about the training results:
Good:
* clay type - I have trained kaolin(white), gray, brown, tan, natural. Using kaolin worked well but colors does not work as well. * position - I can prompt for positions and get really good results. I can prompt for submerged and get a really good result. Standing, sitting, and kneeling works well. * coverage level using weights - I can apply some mud or get her completely covered * multiple or groups of women - this works surprisingly well even though I did not include a single image of two or more women. It will often give them the same face like twins and sometimes create unrealistic limbs. It frequently gets hands and fingers wrong. It creates some really nice images of hugging and touching in mud. * front and back - it is possible to generate images of both sides. * close ups - it can generate close ups of faces or body parts. I did not try feet. * setting - it does not force the setting to be outdoor like in a forest. It needs improvements to avoid altering indoor settings. * mixing substances in mud - it is possible to mix other substances which already work well without training into mud. * hands - hands work well when they are covered in mud and you do not need to worry about the number of fingers right. * hair - I can get mud applied to the hair, but it needs a ton of improvement.
Bad:
* facial - turns faces into witches, pointy chin, goofy eyes, very ugly. I tried training on muddy faces, but personally I like clean faces. * mud colors - this will change the background color (e.g. brown mud makes a brown bathroom) * red clay - it is not possible to generate * specific coverage on areas - it is very difficult to control which parts of the body gets covered and which stay clean. * clothing - if you want her to wear a blue dress covered in mud, it will instead turn her dress into blue mud. It does not understand the concept of applying mud on ( * clothing. Almost all of the images I used in training were fully nude women. * submerged level - like up to the knees, up to the crotch, up the the waist, up to the chest, up to the neck. * eyes - every image I generate will have her looking down with her eyes mostly closed. I need to add to the prompt to make her look at the camera. If she is still * * * looking down, I need to increase the weight of the prompt to make her look at the camera which will make her eyes bigger or unrealistic. * texture - most texture comes out very smooth and uniform. The majority of images were trained on prepared clay. It is hard to incorporate natural mud texture and colors. * laying - I have a difficult time getting good laying positions. * specific actions - like rubbing a clump of mud on her chest or dumping some on her head. * exotic positions - like head dunking, face down ass up, face submerged but not the body, etc. * sexually explicit content - I did not train on any data that includes masturbation or penetration * specific multiple girl actions - it is pretty good with hugging or kissing without having trained on it, but you will not be able to get results for specific things * lesbian - I did not train on any women on women photos. * male - I did not train on male content. It is possible to get males muddy, but I don't know if it will generate genitals. * clean off stage - I did not train on the clean off stage which includes wiping off or showering. * clothing over mod - I did not train on the concept of like putting a dress on over a muddy body. * water - I did not train on concepts involving water like sitting in a lake while rinsing mud off hair * bubbles - It is not possible to generate an image of a bubble in mud * feet - I only included a couple images in training. * sexual body parts - There are some images of anus, vulva, and nipples. It is possible to use these in the negative filter to limit them but it is hard to completely * exclude them. It is possible to include them in the positive prompt. * quicksand - I did not train on quicksand images. * sinking - I did not train on the quicksand idea of sinking where only the subjects face is visible * peat - I did not (at the time) have any peat to train on. * limbs - there are a few photos I used where the arm is obscured in a way which looks like it ends at the elbow. This shows up.
Other:
People - I did not caption the names of the models. It is not possible to reproduce the likeliness of any one particular model intentionally.
It can generate pairs or groups. They all will have the same face. It is possible to use inpainting to give them each a different face but it would take a long time.
The female astronaut seemed a bit too young and fake.
I don't know if I should generate controversial things like a woman being arrested for being muddy and naked in public. I couldn't get the mud on her to look natural. Changes I make to the prompt might unintentionally make the police officer naked or muddy when I want the uniform.
I could not create a decent yoga scene or yoga poses.
The woman sitting with a brick wall is supposed to be in jail and hand cuffed. I could not generate hand cuffs.
The woman sitting on the desk had an amazing pose, but I could not change the mud type without ruining it.
Glasses were difficult. This was as good as I could get.
It is a lot of fun to mix substances. This result has a lot of imperfections, but it just looks like it feels amazing. The results are quite random. I did not train it on applying whipped cream on her body like that. I only trained on mud photography.
I'm interesting in what everything thinks. Any particular concepts which interest you?
paintandmud asks in the photo of the group of women if I could get full coverage including the face. The answer is yes, sort of.
I got some really bad images of faces when applying mud to them. I mean disturbing. Beard hair appears. Disturbing eyes, teeth, and mouths. I think "facial" and "hair" when used in the same prompt will generate facial hair, so using "facial" in my training captions was a bad choice.
Here is the same group shot when I apply terms in the prompt to get some mud on her face. I didn't spend much time fine tuning these. The group photo changes significantly, and they become deeply submerged in the mud. A few close up shots of facial coverage as well as a standing pose of complete coverage. Here eyes were closed. I tried to get her to open them, but I couldn't. I would need to improve the captions on the training to get the engulfed look on the face.
The other problem is that the Restore Face feature will not work correctly. It will blend the mud into her face and lose texture. So these examples have it turned off. You can tell by the teeth and the pupils.
Here is what happens if I try to use clothing. The clothing becomes muddy looking and turns into colored mud. The model I created was only trained on nude images.
I have a feeling like training with clothing will become very difficult. It can be hard to tell where the clothing ends and mud covered skin shows. It can even be hard to tell if the model is wearing panties. There are a lot of types of clothing, and each type needs to be captioned. It will be hard to caption and include terms that can control how messy the clothing is.
Both the speed of the advancement and the potential in the field, as illustrated by your dedicated work here is...stunning. I'm not sure quite what else to say, but - wow.
So much for my retirement plan to produce mud shoots, though.
Plonk said: Both the speed of the advancement and the potential in the field, as illustrated by your dedicated work here is...stunning. I'm not sure quite what else to say, but - wow.
So much for my retirement plan to produce mud shoots, though.
Haha please don't give up on your dream of some mud photography. I only wish I could supply an endless amount of ideas for some real photography and videos. Since I started this, I am starting to imagine new and wild ideas that I have never thought of before. Things that I want to order customs for which I have never done. I would love to quit my desk job and be able to become a photographer too, but I am just too shy to consider it. I have been lurking in umd for years but was too afraid to post anything until I discovered this.
I thought they were real photos at first! Great graphic work! I liked the singer the most. I even imagined myself standing directly on stage at a concert with dirt dripping off the singer's naked body, wow. I liked the photos of pure nudism more than in clothes. Still, you're right that under the dirt is unclear whether the model has panties or not. I would love to see more of your work. You're intrigued by the fact that you have lots of wild ideas.
First of all, thank you. You've gone and done what I was only beginning to study. At least I was headed down the right path.
Most of the photos you used came from my camera. I do confess to some mixed feelings about this. Certainly would not have said "no" to the idea but it might have been nice if you'd asked just out of courtesy. Many of the images are out there for the taking anyway. What's wildest is I can recognize so much of MPV and yet, they aren't. It's a brave new world we're tinkering with here.
Anyhow, very nice work. And thank you for explaining what's going on in detail. I'd love to discuss some things on the side with you.
DuncanEdwards said: First of all, thank you. You've gone and done what I was only beginning to study. At least I was headed down the right path.
Most of the photos you used came from my camera. I do confess to some mixed feelings about this. Certainly would not have said "no" to the idea but it might have been nice if you'd asked just out of courtesy. Many of the images are out there for the taking anyway. What's wildest is I can recognize so much of MPV and yet, they aren't. It's a brave new world we're tinkering with here.
Anyhow, very nice work. And thank you for explaining what's going on in detail. I'd love to discuss some things on the side with you.
Hello DuncanEdwards
Thank you for the compliment. I have enjoyed the photography for years. Sorry I wasn't really familiar with everyone here and didn't know it was your work. I only started participating socially on this forum a few weeks ago. I did email dlodoski but I didn't get a reply. Now I feel like I should have searched and found you first and asked.
I can imagine every artist in the world right now feels threatened that their work will be ripped off or mocked. Even my profile pic of something resembling Andy Worhol's pop art is an example of that. I mean every artist including the photographers, cinematographers, writers, models, producers, etc.
If you were starting to study this, I would be happy to share details on how I trained the model, the images I used, the captions I used, the tools, and the training parameters. I would be happy to discuss the future of how this could be used too.
The photos belong to MPV so attempting to contact Dave was all you could do. Ethics is kind of dead to me, it's just about legalities. Neither of which are defined at all for this yet. Nor are they likely to be anytime soon. The real world is that someone would have done, will do, unspeakable things with all the content that is out there.
Please contact me directly at Duncan.Edwards036@gmail.com when convenient. I'd really like to learn a little more. Maybe we can help each other out.
DuncanEdwards said: The photos belong to MPV so attempting to contact Dave was all you could do. Ethics is kind of dead to me, it's just about legalities. Neither of which are defined at all for this yet. Nor are they likely to be anytime soon. The real world is that someone would have done, will do, unspeakable things with all the content that is out there.
Please contact me directly at Duncan.Edwards036@gmail.com when convenient. I'd really like to learn a little more. Maybe we can help each other out.
I feel really upset that you think ethics is dead. I'm really sorry. If it was, or will be, illegal for me to use the photos for training, I'm really sorry and I could delete this post. Maybe this post would encourage others to train illegally on photos and do bad things with it.
I'm thinking maybe it would be best to ban generated images from this site all together. It would avoid all of the problems.
Right now the only thing holding people back is that most online generators have filters. Once it becomes more accessible, I guarantee the amount of generated images on this site will increase, and it will probably just bring unspeakable things.
Maybe this sort of content doesn't belong here and banning it would be better. People can host and look for it elsewhere. I really like real content and I don't want that to change.
I'm sorry I might contact you next week because I have been feeling pretty bad about what I have done and need to think whether or not I want to continue.
I've flagged this thread to MM to review. My personal take (as I also mentioned on the "Ethics" thread) is that while it's clearly impossible to stop people doing whatever they like with content they have purchased, we should ban the posting, here, of content generated by a model trained on *actual wam models*, as otherwise it's a wide open door for mass piracy using effectively clones of known people.
Generic fantasy people the AI creates, fine. Almost perfect liknesses of real people who are part of our community, without their consent? Hard no.
DungeonMasterOne said: I've flagged this thread to MM to review. My personal take (as I also mentioned on the "Ethics" thread) is that while it's clearly impossible to stop people doing whatever they like with content they have purchased, we should ban the posting, here, of content generated by a model trained on *actual wam models*, as otherwise it's a wide open door for mass piracy using effectively clones of known people.
Generic fantasy people the AI creates, fine. Almost perfect liknesses of real people who are part of our community, without their consent? Hard no.
Thank you DungeonMasterOne.
It was certainly not my intent to generate the likeliness of a specific MPV model. It is not possible to produce a single MPV model with the training I did because I did not caption the names of the models in the training, but I found some of the images has aspects which resembles the MPV models.
It was not my intent to generate images that plagiarises real MPV photography. I wanted to include more images from other producers in the training to avoid this, and I really only wanted to make creative images that you would never see like the first examples.
From 1 week after I posted this I went from feeling proud to ashamed. If generated images needs to be banned to save umd then let it be. I would rather feel guilty that my post led to it being banned. If it isn't banned, then then I might still continue posting images and I will do so respectfully. I was interested in training on other wam concepts other than mud like slimes and pies. This just has me feeling really conflicted though so I wasn't sure if I should continue.
Let me also say that the trained model which I produced is 144 MB file. If this is shared, it gives anyone the capability to produce images in the same manner which I did. I will not share it. I am even concerned sharing details on how I trained it. There are websites where people share these models. I don't think it is piracy, but it certainly could have the same impact. The generated images can certainly be plagiarism is the images generated look exactly like the producers style like I did in a few of them.
Anyone with a good enough GPU and who can follow a youtube tutorial can do this. I started on April 16th. I trained 3-4 other things for practice. I finished training on May 28th which is when I posted the ethics thread. Hopefully this information helps when making decisions.
Some more examples of some fun I am having with this. I'm still conflicted, but it is so much fun to generate and share this.
* Canoer who found an interesting way to enjoy the ride. * Wanderer who astounded by her discovery of a a secret flower garden and natural mud spring. * The restaurant owner arranged a promotion giving customers a special view on the opening day. * Rafter who wants her hero to rescue her at sea * Golfer who follows strict course rules when the ball falls into the mud pit.
The restaurant scene was originally a mcdonalds. It included the logo. I wasn't sure if I should post it, so I tweaked it. This variant is not as good, but the indoor scene looked pretty good. The mcdonalds variant had a better facial expression which was slightly naughty.
I have thousands of similar images. I select a dozen or 2 of the ones I like, then select the best one to try and tweak and upscale. Some ideas are very difficult.
mFeelzGood said: Some more examples of some fun I am having with this. I'm still conflicted, but it is so much fun to generate and share this.
* Canoer who found an interesting way to enjoy the ride. * Wanderer who astounded by her discovery of a a secret flower garden and natural mud spring.
Okay, try not to get all freaked out about what I'm going to say because I'm really enjoying what you are doing. It's a great learning experience. I'm looking at your images and often I see my work, the locations we used, the ladies we photographed, but at the same time they aren't any of those things. I don't know exactly what to call it but there's a style or a theme here that I recognize. If I saw some of these photos elsewhere I'd be trying to remember when I shot it. This is fascinating.
I had a photography teacher once and someone asked him how we develop a style or form of our own. His answer was great - "You don't have any style. Not yet. Look at someone else's work that you like and copy it. Then make it yours."
I feel like that's exactly what's taking place here. It's not wrong or unethical or immoral or messed up or anything. Can AI make it's own style or does that fall strictly to the human element? Can a human who may not be endowed with any particular creative skill properly use what AI has learned from someone who knew what they were doing? We're going to get into some interesting questions here. I just don't know what many of those questions might be yet.
DuncanEdwards said: Okay, try not to get all freaked out about what I'm going to say because I'm really enjoying what you are doing. It's a great learning experience. I'm looking at your images and often I see my work, the locations we used, the ladies we photographed, but at the same time they aren't any of those things. I don't know exactly what to call it but there's a style or a theme here that I recognize. If I saw some of these photos elsewhere I'd be trying to remember when I shot it. This is fascinating.
I had a photography teacher once and someone asked him how we develop a style or form of our own. His answer was great - "You don't have any style. Not yet. Look at someone else's work that you like and copy it. Then make it yours."
I feel like that's exactly what's taking place here. It's not wrong or unethical or immoral or messed up or anything. Can AI make it's own style or does that fall strictly to the human element? Can a human who may not be endowed with any particular creative skill properly use what AI has learned from someone who knew what they were doing? We're going to get into some interesting questions here. I just don't know what many of those questions might be yet.
It is really cool that you can recognize your style. Even someone who commented on a photo recognized the style.
I have seen people apply artwork styles. I tried to apply photography styles from some photographers, and it didn't have much of an affect. It changed the facial expression and color, but not the body, pose, or background. I'm not sure what to prompt for to try different photography styles.
I am working on an experiment. I want to train on Muddygirlies photography. I have 600 images that I need to caption. I want to see if it will reproduce images which looks like Muddygirles' style. Then I'm going to try to train a model with both 600 Muddygirlies photos and 600 MPV photos to see how it combines them. I think training one one producer will copy the style of the producer, and merging them would produce recognizable aspects of both producers instead of creating something new.
I don't have any creative talent, so this tool plus your photography is what enables creativity. I think the quality of the images makes a huge impact on the end result.
For the first attempt, I generated an image where the subject was nude and body covered in mud.
Then I used inpainting to add cloths back on. Then I used inpainting to add mud back and take the cloths back off.
These did not turn out that well. inpainting can look like a botched photoshop. It does work to make tweaks. I had a hard time trying to make her put her hands on her head or change positions. I had a hard time upscaling, so these are not upscaled.
This uses a constant seed, and all I am doing is changing the prompt or weights in the prompt. This adds some random changes to the the position which is nice, but it also changes the background too much.
With inpainting, I can keep things consistent but have a hard time changing things. With this approach, it is good at changing things but introduces a lot of inconsistency. Her breast size change a lot.
Clothing is a huge struggle for me because I did not train with it.
I used inpainting to make the subject clean again. Then I put cloths on her. Then I applied a different type of mud on her. The individual results are far from perfect, but the series is quite interesting to watch.
It is a lot of fun. It is like those days at work where I day dream and wish I could pause time.
1. over 600 512x512 images from MupPuddleVisuals 2. over 600 512x512 images from MuddyGirlies 3. over 1200 512x512 from both combined.
Columns: 1) The original MPV Lora, 2) 600 MPV, 3) 600 MuddyGirlies, 4) 1200 Combined Rows: 1: 3 seeds to compare the 4 models. 2: body poses: A) kneeling, B) Sitting, C) Submerged 3: mud colors: A) Earthen, B) Tannish, C) Grayish 4: mud colors: A) Kaolin, B) Yellowish, C) Brownish 5: hair color: A) Blond, B) Brunette, C) Redhead 6: ethnicity: A) Asian, B) Latino, C) Black 7: mix with slime colors: A) White, B) Green, C) Blue 8: mix with substances: A) Lotion, B) Chocolate, C) Cream 9: facial coverage: A) Little, B) Some, C) More 10: focused body part: A) Boobs, B) Butt, C) Crotch 11: 3 seeds to show concept of hand applying clay mud 12: 3 seeds to show side profile improvement 13: 3 seeds to show engulfed look 14: 3 seeds to show a mix of 3 of my loras including sweaty girls and pottery girls. 15: setting: A) Pit, B) Wallow, C) Quarry 16: background: A) Forest, B) Jungle, C) Lakebed 17: clothing: A) Dress, B) Skirt, C) Bikini 18: hairstyle: A) Long, B) Bobbed, C) Pixie 19: 3 seeds to show bathtub 20: 3 seeds to compare toonyou stylized 21: 3 seeds to compare animesh stylized 22-25: Upscaled experiments with mixing chocolate syrup, whipped cream, and mud concepts together. This is using the combined model. 26-27: Upscaled experiments with mixing a blue slime on mud. This is using the combined model. 28: Upscaled experiment adding orange flower petals. This is using the combined model.
Everything is cherry picked to show the better results and remove most results with defects.
Summary:
* The first MPV model often results in thick layers and blobs of clay mud because the images I trained on were my favorite ones which featured a lot of this. It is really bad with faces. It will often generate open mouths which I cannot fix. * The new MPV model uses a lot more Muddy Pinay Models, and has a lot more muddy faces. It will often make the faces muddy when I don't want it to. * The first MuddyGirlies model generates beautiful outdoor images, but it will often turn the background into mud. It will mostly generate clean faces, but it is capable of generating muddy faces. * The combined model seems like an average of the 2 and has the capabilities of both. It will generate backgrounds which look like the MPV mud pit, the MuddyGirlies quarry, and mixes of the two. It is really good with mixing concepts.
I tried to train a few more concepts in this model, but it didn't work well. I'm going to work on an improved model.
* 512x1024 resolution * upscaling 2x to 1024x2048 * smiling facial expressions * random hair styles * random mud colors
These side by side images are just changes to weights of smiling from 0.0 to 1.0. I generated 3k of these, select 77 nice ones, and then selected 25 of my favorites to upscale. The background is supposed to be a lake at dusk.
Some interesting things about changing a smile:
* eye colors sometimes changes. I added hazel eyes to the prompt to avoid it. * hair will be the same style but have small details changed. * arm positions will frequently change and leg positions will change rarely. * the are subtle background changes.
About 512x1024:
* the subject will often have a really long body. I included these because the facial expression was really nice. You will just have to imagine that she is 6'3 hehe.
I really like when a smile is captured. Maybe the photographer said something really nice. Or maybe it is the moment where she realizes how much she is enjoying it.
I also really enjoy the look of a mud hand-bra. I did not prompt for this. These just randomly occur. The pink hair ones would often make the dusk sky pink and purple which I really liked.
Some images that I thought of by seeing a post about wam con 2024. The buildings in the background are meant to be hotels. I was thinking about replying in the post about a perfect venue for wam con as a joke, but decided not to because I didn't want to cause confusion.
DuncanEdwards said: Okay, try not to get all freaked out about what I'm going to say because I'm really enjoying what you are doing. It's a great learning experience. I'm looking at your images and often I see my work, the locations we used, the ladies we photographed, but at the same time they aren't any of those things. I don't know exactly what to call it but there's a style or a theme here that I recognize. If I saw some of these photos elsewhere I'd be trying to remember when I shot it. This is fascinating.
I had a photography teacher once and someone asked him how we develop a style or form of our own. His answer was great - "You don't have any style. Not yet. Look at someone else's work that you like and copy it. Then make it yours."
I feel like that's exactly what's taking place here. It's not wrong or unethical or immoral or messed up or anything. Can AI make it's own style or does that fall strictly to the human element? Can a human who may not be endowed with any particular creative skill properly use what AI has learned from someone who knew what they were doing? We're going to get into some interesting questions here. I just don't know what many of those questions might be yet.
From an technical standpoint, AI is not making its own content. It's taking what it sees, manipulating it, and calling it its own. That's no different from what a restaurant chain did to me with some photos I took for one of its employees, and they ended up owing me a fairly large settlement thanks to US copyright law. It's also not a whole lot different from the old sampling battles of the 80s and 90s with music (i.e., Vanilla Ice; interesting read if you're younger and unfamiliar).
Point is, as a photographer and creator, I would be quite unhappy if my work was used without permission for AI learning, and would be demanding compensation if the end result was work being sold that was trained off my photos. Consent is a very smart thing to obtain going forward, else the 'trainer' may be viewed as an IP thief.
That said, mFeelzGood, the work is coming along and what you're doing is cool. But I encourage you to DEFINITELY get consent going forward for any assets you train your tools with, and maybe step outside of just the AI box for a bit. Get a camera or dust off the phone, find a local subject who may be willing to let you shoot some messy content for your own tools' training, and enjoy the creative side of photography rather than just the manipulation of existing imagery. You might find a new passion there.
From an technical standpoint, AI is not making its own content. It's taking what it sees, manipulating it, and calling it its own. That's no different from what a restaurant chain did to me with some photos I took for one of its employees, and they ended up owing me a fairly large settlement thanks to US copyright law. It's also not a whole lot different from the old sampling battles of the 80s and 90s with music (i.e., Vanilla Ice; interesting read if you're younger and unfamiliar).
Point is, as a photographer and creator, I would be quite unhappy if my work was used without permission for AI learning, and would be demanding compensation if the end result was work being sold that was trained off my photos. Consent is a very smart thing to obtain going forward, else the 'trainer' may be viewed as an IP thief.
That said, mFeelzGood, the work is coming along and what you're doing is cool. But I encourage you to DEFINITELY get consent going forward for any assets you train your tools with, and maybe step outside of just the AI box for a bit. Get a camera or dust off the phone, find a local subject who may be willing to let you shoot some messy content for your own tools' training, and enjoy the creative side of photography rather than just the manipulation of existing imagery. You might find a new passion there.
Thanks for your comment
Before my first post on umd, I had the trained model ready. I did email dlodoski. I wanted to show the images that I produced and then get permission to post them. I didn't get a response. My email probably came off as weird or was ignored. I could have tried more to contact him.
I was going to contact the muddygirlies site owner too. I did send an email with a question about the subscription, but I never got a response. I figured that I would not get a response about permission to use his photography as well.
I feel like I would be ignored by majority of producers if I asked for permission. They are probably used to receiving a lot of weird requests and are used to ignoring them. Maybe they would respond if this would benefit or impact them financially.
If dlodoski or muddy contacted me and asked me to stop, I would stop posting images here. If one ore two more producers replied here with the same sentiment where they do not want their photography being used to train models and that it is a form of IP thievery, I would stop posting images here. I can understand why producers would want me to stop because it took them a lot of effort and setup to produce the photography, and then they see someone like me who posts mimics of their hard work.
If producers would demand for compensation for their photography being used for training, I would stop. I don't intend on selling or taking commissions to make these images. I'm just doing this for fun to see what images get loved and for the comments.
If copyright law is changed to make training no longer applicable for fair use for the content that I purchased, then I will stop.
I would love to produce my own photography to train on. I don't think I would ever have the courage or social skills to do it. It would also be very difficult to set up a mud pit like in the MPV set, or go out to a remote location like a quarry. It would cost me more than I would every make, and I would risk being rejected by the community here for producing photography which is similar to MPV and Muddygirlies. I probably wouldn't be able to create my own style, and it would just look like copies.
The sexiest thing about dirty pictures is knowing that the subject willingly got that dirty and enjoyed it, making it feel like there's a chance of someday finding someone who'll do that with you.
This is similar to pornography, to some extent, in that when the subject is actually having a good time and it's not just a performance for the camera, it is so much better and real world relatable.
I can't really enjoy AI pictures because I know they're not real, as it is the real person being naughty that is the actual turn on not just the imagery on offer. I'm sure there's plenty of other who do, but I want to see real women who enjoy getting dirty, so I know that lots of folk are enjoying filthy play together.
Muddddy said: The sexiest thing about dirty pictures is knowing that the subject willingly got that dirty and enjoyed it, making it feel like there's a chance of someday finding someone who'll do that with you.
This is similar to pornography, to some extent, in that when the subject is actually having a good time and it's not just a performance for the camera, it is so much better and real world relatable.
I can't really enjoy AI pictures because I know they're not real, as it is the real person being naughty that is the actual turn on not just the imagery on offer. I'm sure there's plenty of other who do, but I want to see real women who enjoy getting dirty, so I know that lots of folk are enjoying filthy play together.
I couldn't agree more. Knowing that an experience was real is what makes it enjoyable to view. This is what makes AI images disappointing. It creates a fake experience. WAM is about the experience and feeling.
People tend to generate images of their kinks, and that is what makes AI images more enjoyable than real images for them. For me, I cannot enjoy something where the person is not enjoying it. I can only enjoy it if I feel a connection that the other person enjoyed it.
I have never been muddy. I have only imagined it. To me, WAM has always been an imagination or daydream experience. So it is quite fun for me to generate images from my imagination right now.
I hope that AI images and video does not replace real content. I hope it is only used to inspire ideas for real content. I hope it does not do the opposite where producers see AI images and lose motivation for creating content.