Cool video! I think broadsword might of been too specific, just a plain sword might have had better results. I’d be curious to see what it could do with a bejeweled dagger.
I use Midjourney to make character profiles. I find that it helps to give the AI a reference for who the person should look like. Particularly with nale Portraits it seema to struggle with making a good face unless it has a reference. It also needs to know things like hair color and clothing color. Eye color doesn't work right at all so don't bother. I then fix it up in Photoshop with a combo of liquify copy and pasting the best eye and flipping (one eye is ALWAYS crappy) to replace the bad eye, hand painting, replaying backgrounds when needed and generally touching things up. Also, using --uplight in the prompt helps to get a cleaner looking face.
I have a premium tier MidJourney account and, whilst it might take a few iterations on descriptions/command words, it works incredibly well. You have to play around with your wordings, thinking about the image in the sense of emotions/moods to pull the final image in the direction you want. Variations will only make very close variations of the same image, if it's not spitting out what you want, add or subtract words. The results you can get are mind-blowing and, whilst lacking ultimate specificity, beats what most artists could churn out in anything less than a few day's work, if they're really good. As long as you don't want something very precise, like an exact image of your character/landscape in your mind, you'll produce something really good with minimal effort.
I can see how a tool like AI generator could be a good starting point to some people, and help non-artists with image ideations and such. What I find most troublesome is how most of these generators feed artworks from currently living professional artists that didn't give permission to this. Like the example with having James Gurney as an option, while he is alive, well and still working. As a professional artist I feel like these and NFTs are punishing those of us that choose to publish our artwork online.
I've seen a couple AI generated pieces of fantasy art where you can see in the bottom corner it's fused together a number of artist's watermarks/signatures. Definitely concerning times.
@@miljantrajkovic1862 inspiration is one thing, a carbon copy is another. This is already a problem with humans, I just hope there's a way for artists to opt out or protect themselves, other than just exiting the internet.
I think your issue with Midjourney or Midjourney Variations as a whole comes from not understanding the mechanics. Variations basically run the same prompt using the selected image as a refrence point, so it basically is assuming you like that composition and want the same elements but using the prompt again with that selected image as a base. Typically you want to tweak your written prompts off your refrence image of choice otherwise you will have to run multiple iterations before getting something that fits what you want
With MidJourney You can add in a whole bunch more detail and commands to get the image that you’re really looking for. I’ve been messing around with it for a little while now and have absolutely loved it. He does have problems trying hands and sometimes even messes up the face but some of those things can actually be touched up with On an image manipulation program.
Really like the sword end product for mid journey yea it’s not exactly what you asked for but still think it turned out a pretty cool Excalibur-esque product
My brother showed me this technology and it's great if you have a refined description, know what you are doing and have some art skill with lots of time.
I've been creating art with Midjourney and I have got some decent results with it. Obviously the more specific you need the image to be the more frustrating it can become. Knowing keywords for styles and the like is incredibly useful as I wanted most of my art to look alike. There are details that are lacking, as Midjourney seems to really struggle in creating eyes that look the same, which is why it is probably good to use photoshop to tweak these images to make them actually useable. I plan on using Midjourney to generate basic landscapes and simple characters. This allows me to stretch my budget and pay for quality art by getting real artists to do the kind of art that these AI bots simply cannot handle.
I find that creating portraits in Midjourney and then importing them into Artbreeder creates great results. At least for humans. Artbreeder knows what eyes and ears should look like.
@@LunchBreakHeroes I'll definitely have to give this a try sometime. I have used Srtbreeder in the past but it is a tricky program to use to make art from scratch so this might be a good middle ground.
fun but honestly a 3D render is the way to go if it were me, lots of free options like Unreal Engine with Quixel Megascans included, so many freebies each month too
What inspired this video by the way? I recall a year ago some friends were playing a guessing game of film and TV titles with abstract art similar to night care and starry that had been created by an AI using the plot synopses from imdb; were some wild images.
I use it. Not because I love it or feel it's superior to human created work, but because my release schedule is so tight that there's no way for a commissioned artist to create the imagery I'd like to showcase. Whereas multi-employee publishers like MCDM work on a timeline where their stuff is planned and worked on months ahead of release, me myself and I are turning things around by the seat of my pants in a matter of weeks or days. That being said, for printed releases like Librus Nocturnum (and future published books whose timeline is measured in months/years), AI artwork is removed and replaced with commissioned artwork.
@@LunchBreakHeroes thank you for the insight. Makes sense as I’ve been exploring different kinds of options. I agree with bigger releases like Kickstarter projects have the time and capital for it. How’s the recent Kickstarter project going?
MidJourney is pure magic to me. With more exact commands you can create just about anything with incredible quality. With good instructions and understanding how it works you can recreate just about any fictional moment you can imagine.
Hey, that Guardians of the Galaxy joke was pretty great! What do you call the security guards standing outside of an Iphone factory? Suicide Squad. You know, because of all the workers who committed suicide due to horrendous working conditions. heh. Ok, not funny.
I think Dall-E would have performed better than Midjourney. I generated a ton of imaged for my CoS game with it and its quality really blew me away. Wish I had more credits because I used them all up in an hour or two!
Wildly misinformed about how AI art generators work. They do NOT search the internet looking for images that have something to do with your prompt. They can be used offline with a local installation. This kind of misinformation contributes to the major uninformed backlash AI art gets online. Took less than 3 minutes to get my 👎🏻. Edit: Giving Midjourney 4-8 tries and an upscale, vs one overbaked attempt for the other two also drastically biases your results.
Use mods to make your Skyrim character look like your D&D character.
Take screenshot.
Apply style transfer.
Free art.
Cool video! I think broadsword might of been too specific, just a plain sword might have had better results. I’d be curious to see what it could do with a bejeweled dagger.
I use Midjourney to make character profiles. I find that it helps to give the AI a reference for who the person should look like. Particularly with nale Portraits it seema to struggle with making a good face unless it has a reference. It also needs to know things like hair color and clothing color. Eye color doesn't work right at all so don't bother. I then fix it up in Photoshop with a combo of liquify copy and pasting the best eye and flipping (one eye is ALWAYS crappy) to replace the bad eye, hand painting, replaying backgrounds when needed and generally touching things up. Also, using --uplight in the prompt helps to get a cleaner looking face.
I've used Midjourney and Night Cafe, and have found that the art created really depends on how the promt is written.
Guardians of the Galaxy. Hilarious. Yes, I am a dad🤣
I have a premium tier MidJourney account and, whilst it might take a few iterations on descriptions/command words, it works incredibly well. You have to play around with your wordings, thinking about the image in the sense of emotions/moods to pull the final image in the direction you want. Variations will only make very close variations of the same image, if it's not spitting out what you want, add or subtract words. The results you can get are mind-blowing and, whilst lacking ultimate specificity, beats what most artists could churn out in anything less than a few day's work, if they're really good.
As long as you don't want something very precise, like an exact image of your character/landscape in your mind, you'll produce something really good with minimal effort.
Midjourney is in a whole different weight class. A fairer competition would have been against Dalle2 , Stable Difussion or maybe even Disco Difussion.
Didn’t get into Dall-E in time, and haven’t heard of the other two
@@LunchBreakHeroes Ah, makes sense. Still cool video, through the other two never even stood a chance honestly 😅
I can see how a tool like AI generator could be a good starting point to some people, and help non-artists with image ideations and such. What I find most troublesome is how most of these generators feed artworks from currently living professional artists that didn't give permission to this. Like the example with having James Gurney as an option, while he is alive, well and still working.
As a professional artist I feel like these and NFTs are punishing those of us that choose to publish our artwork online.
I've seen a couple AI generated pieces of fantasy art where you can see in the bottom corner it's fused together a number of artist's watermarks/signatures. Definitely concerning times.
But it's same as human intelligence, we take inspiration from other artists.
@@miljantrajkovic1862 inspiration is one thing, a carbon copy is another. This is already a problem with humans, I just hope there's a way for artists to opt out or protect themselves, other than just exiting the internet.
@@Lunaishtar The AI doesn't make a carbon copy of someone else's artwork. It mimics styles, but it doesn't copy the actual art pieces itself.
how do NFTS punish you xD , nfts only prove authenticity and proof of origin , it is bad actors stealing art thats affecting you
I think your issue with Midjourney or Midjourney Variations as a whole comes from not understanding the mechanics. Variations basically run the same prompt using the selected image as a refrence point, so it basically is assuming you like that composition and want the same elements but using the prompt again with that selected image as a base. Typically you want to tweak your written prompts off your refrence image of choice otherwise you will have to run multiple iterations before getting something that fits what you want
With MidJourney You can add in a whole bunch more detail and commands to get the image that you’re really looking for. I’ve been messing around with it for a little while now and have absolutely loved it. He does have problems trying hands and sometimes even messes up the face but some of those things can actually be touched up with On an image manipulation program.
Really like the sword end product for mid journey yea it’s not exactly what you asked for but still think it turned out a pretty cool Excalibur-esque product
Definitely a very cool result!
My brother showed me this technology and it's great if you have a refined description, know what you are doing and have some art skill with lots of time.
Thank you for your passion!
I've been creating art with Midjourney and I have got some decent results with it. Obviously the more specific you need the image to be the more frustrating it can become. Knowing keywords for styles and the like is incredibly useful as I wanted most of my art to look alike. There are details that are lacking, as Midjourney seems to really struggle in creating eyes that look the same, which is why it is probably good to use photoshop to tweak these images to make them actually useable. I plan on using Midjourney to generate basic landscapes and simple characters. This allows me to stretch my budget and pay for quality art by getting real artists to do the kind of art that these AI bots simply cannot handle.
I find that creating portraits in Midjourney and then importing them into Artbreeder creates great results. At least for humans. Artbreeder knows what eyes and ears should look like.
@@LunchBreakHeroes I'll definitely have to give this a try sometime. I have used Srtbreeder in the past but it is a tricky program to use to make art from scratch so this might be a good middle ground.
fun but honestly a 3D render is the way to go if it were me, lots of free options like Unreal Engine with Quixel Megascans included, so many freebies each month too
So what licensing rights do you get from the midjourney images? Can you, as a subscriber, get images that you own and are able to publish?
Yup, you can use them no problem.
You have the rights, but the art isn't copyrighted, so anyone can copy your work without a problem.
Many of these ais can't use stuff like tiefling in their training for copyright reasons
This is incredibly useful! Thank you for this video. Anyone know which of these tools has the best policy as far as use for actual plays, etc?
Would've liked to see Dall-E involved as well
I applied to their beta but didn’t get in in time!
@@LunchBreakHeroes I have the beta! If you'd like, I can run some screen capture on whatever prompts you like and see what we get?
Feel free to run these same prompts through and share them to our Discord. I’d love to see them!
What inspired this video by the way?
I recall a year ago some friends were playing a guessing game of film and TV titles with abstract art similar to night care and starry that had been created by an AI using the plot synopses from imdb; were some wild images.
How would you approach using AI art in your Patreon creations and adventures?
I use it. Not because I love it or feel it's superior to human created work, but because my release schedule is so tight that there's no way for a commissioned artist to create the imagery I'd like to showcase. Whereas multi-employee publishers like MCDM work on a timeline where their stuff is planned and worked on months ahead of release, me myself and I are turning things around by the seat of my pants in a matter of weeks or days. That being said, for printed releases like Librus Nocturnum (and future published books whose timeline is measured in months/years), AI artwork is removed and replaced with commissioned artwork.
@@LunchBreakHeroes thank you for the insight. Makes sense as I’ve been exploring different kinds of options. I agree with bigger releases like Kickstarter projects have the time and capital for it. How’s the recent Kickstarter project going?
It looks like the 2 and 3rd program share the same program
trying to get a red tiefling on midjourney is a pain
MidJourney is pure magic to me. With more exact commands you can create just about anything with incredible quality. With good instructions and understanding how it works you can recreate just about any fictional moment you can imagine.
Scary that an AI created those images... what will AI do next...
Cool video idea; definitely different and unique and interesting.
Thank you! Glad you enjoyed it 😀
Next step is going to be Accurate voices that can almost be indistinguishable from a real person. which we are already half way there.
Really interesting video!
Hey, that Guardians of the Galaxy joke was pretty great!
What do you call the security guards standing outside of an Iphone factory? Suicide Squad. You know, because of all the workers who committed suicide due to horrendous working conditions.
heh. Ok, not funny.
I'm pretty sure you threw it off when you said "Lovecraftian"
I think Dall-E would have performed better than Midjourney. I generated a ton of imaged for my CoS game with it and its quality really blew me away. Wish I had more credits because I used them all up in an hour or two!
Having acces to both midjourney and dall-e I think dalle does better portraits, but midjourney does better landscapes.
Wildly misinformed about how AI art generators work. They do NOT search the internet looking for images that have something to do with your prompt. They can be used offline with a local installation. This kind of misinformation contributes to the major uninformed backlash AI art gets online. Took less than 3 minutes to get my 👎🏻.
Edit: Giving Midjourney 4-8 tries and an upscale, vs one overbaked attempt for the other two also drastically biases your results.
So lets just use an ai that steals the art from all artists that charge money, and get it for free
AI is art theft
Some things shouldnt be done, although they can be done.
Humans Need Not Apply.