Honestly I'm surprised GRRM or another fantasy writer of his level hasn't hire you as a map maker yet!
Your stuff is much better than the official maps I see in ASOIAF.
Your stuff is much better than the official maps I see in ASOIAF.
Artwork Culsu's Art & Maps
- Thread starter Culsu
- Start date
Compensation is... well, hopefully adequate most the time. There's a lot of competition in this little niche, and if you're doing the fully handdrawn stuff like I do you obviously can't compete with the prices that some Pakistani guy on Fiverr throws out for maps he's doing on Inkarnate or Wonderdraft with no or hardly any additions.I hope that you got generously compensated for your work! I myself could unfortunately never personally make maps as beautiful as these ones.
I do manage to schedule/plan most maps in a way that I end up above the comparable local minimum hourly wage, luckily.
As for you no being able to make something like my maps: honestly, ask me ten years ago and I would have said the same. What you see in this thread is, above anything else, the result of patience rather than inate skill. So, my advice to you? Give it a try!
Huh, I was prepared to tell you that the guys and gals working for the big hitters are way better than me, but then I looked at Jonathan Roberts who did the maps for the GRRM compendium, and whose tutorials I used frequently in my earlier mapmaking days... and I think I don't have to hide my maps when compared to his (his GoT city maps are still way out of my league, though!).
That being said, most authors contracted by large publishing houses have little to no say in what kind of map artist does their stuff, which is why you get so many fantasy novels either using the authors' own (poor) sketches or using something that looks as if it's been slapped together in three hours by someone with a mouse in MS Paint: because it's cheap.
I've been doing this as an amateur, and now professionally, for close to ten years, which is why I can gauge how much effort any given map will take pretty well. Black & white maps? Around ten hourse, give or take a couple. Colored? Twenty plus. Adjust either for size and extra details, and you can get up to fifteen for the black & white maps, and... pretty much open ended for the colored ones. Longest time I ever sat at a map? Probably fifty hours.
Big publishers are ripe with horror stories of ripping off their creatives. Latest case in point. I guess what I'm trying to say is this: know your worth. Know what your time is worth. Don't work for name recognition, or to get a "big" industry name on your list of former clients, because nothing will ever come from it. Work the jobs that pay. If Random House or Marvel sent me an email tomorrow with a commission offer that amounted to the same stuff that Anna mentioned in the linked video? That'd either land in my trash folder directly or earn them a barely polite reply pointing out my rates and the required workload.
If you dig through all the horror stories from creatives a pattern will emerge, that pattern being that more often than not, the bigger the publisher, the shittier the pay. So I'd much rather work with indies and small publishers where things are negotiated fairly and up front, and if you don't find a middle ground, well that's life.
Last edited:
The Wending Coast
Map WIP for a game with big stompy robots
Well, the irony. I had finally gotten around to start experimenting with battle maps for Battletech when the whole Blain Pardoe debacle's exploded onto the internet.
Well, maybe as an effort to show that you don't need official products to enjoy an IP here's a WIP of my very first battlemap usable for games with big stompy robots. It's not finished yet, but the basics are remarkably easy to establish once you've figured them out. Truth be told, the most time-consuming part is creating and/or searching for fitting textures.
So, withhout further ado, here it is. Always remember: it's a WIP!
Well, maybe as an effort to show that you don't need official products to enjoy an IP here's a WIP of my very first battlemap usable for games with big stompy robots. It's not finished yet, but the basics are remarkably easy to establish once you've figured them out. Truth be told, the most time-consuming part is creating and/or searching for fitting textures.
So, withhout further ado, here it is. Always remember: it's a WIP!
Wooded Hills Ambush HexMap (Battletech-compatible)
Okay, here it is. I've refined it The map is explicitly compatible with the official hex sizes and nomenclature of BATTLETECH, but can also be freely used in any other context where you have to stage an encounter in such an environment. The map size is 18 x 22 inches (ie. 15 x 17 hexes).
I'm sorry for the watermarks, but you can get the clean file here on a Pay-What-You-Want base, meaning you don't have to pay for it if you don't want to (not that I'd complain if you left me a few shekels and/or a constructive review).
EDIT: I also looked at the official BTech map stuff, and aside from one exception I'm less than impressed. For example, I hope nobody cons you into paying $1.99 for this. That's 15 minutes of work, on a bad day. That's just no value for your money.
I'm sorry for the watermarks, but you can get the clean file here on a Pay-What-You-Want base, meaning you don't have to pay for it if you don't want to (not that I'd complain if you left me a few shekels and/or a constructive review).
EDIT: I also looked at the official BTech map stuff, and aside from one exception I'm less than impressed. For example, I hope nobody cons you into paying $1.99 for this. That's 15 minutes of work, on a bad day. That's just no value for your money.
Last edited:
Rolling Woodlands (HexMap 36x48)
Rolling Woodlands expands on the smaller pay-what-you-want map Wooded Hills Ambush, quadruppling the available size and turning it into one very large (36x48 inches) hex battlemap consisting of woodlands of variýing density, hills and cliffs of varying heights, and terrain of varying difficulty. The map is explicitly compatible with the official hex sizes and nomenclature of BATTLETECH, but can also be freely used in any other context where you want to play a prolongued battle in varied terrain with plenty of room for cover and maneuvering.
The map size is, again 36 x 48 inches (ie. 15 x 17 hexes) and is provided as print-ready (300 dpi) PDFs file as well as high quality JPEGs for your convenience. You get the complete map file, and the map in four parts for easier printing as well.
This one isn't free as it took me a whole lot if time and effort to get done, but $3.99 seems a fair price for a poster-sized battlemap to me. You can get it at DriveThruRPG.
The map size is, again 36 x 48 inches (ie. 15 x 17 hexes) and is provided as print-ready (300 dpi) PDFs file as well as high quality JPEGs for your convenience. You get the complete map file, and the map in four parts for easier printing as well.
This one isn't free as it took me a whole lot if time and effort to get done, but $3.99 seems a fair price for a poster-sized battlemap to me. You can get it at DriveThruRPG.
Huh. I believe these central hills are pretty good for taking control of the battlefield, aren't they?
I'm going to let you in on a secret: I have no clue about the Battletech tabletop rules.
But yes, purely by logic they should give the party holding them an edge, though forests provide cover boni, afail (going by the PC game), so they aren't a direct "I win" card.
Bear Ribs
Well-known member
Woods provide cover which gives an edge, but they also make movement slower which removes bonuses for moving fast. Ideally, you want to spend all your movement in a straight dash up until your last one that goes into the woods, but woods inconsiderately don't always grow in perfectly positioned 1-hex patches.
I'm really digging how varied this map is though, especially the lake, way too little water in most BT maps for my taste. One of the reasons boats tend to be never-used is because there're almost no maps with rivers, coasts, and lakes for them to operate in.
I'm really digging how varied this map is though, especially the lake, way too little water in most BT maps for my taste. One of the reasons boats tend to be never-used is because there're almost no maps with rivers, coasts, and lakes for them to operate in.
I didn't want to have it look all samey. There are way too many official maps on DTRPG that are just utterly lazy (leaving out those from Kat Wylder). This here, for example, costs $1.99 and would take anybody only tangentially familiar with GIMP, Krita or PS probably all of 20 minutes to draw.
Had DiscoDiffusion running in the back while I was doing work on some maps. DiscoDiffusion is one of several popular "AI" art generators that are being hyped from here to hell and back at the moment. Honest to god's truth? After playing around with the free available runtime I give it a resounding 'Meh'. This down here was made with the prompts "Howard Shore landscape, photorealistic, ruined castle on wooded hill, artstation, detailed, unreal engine, fog" and sadly was the last piece I could have created before free GPU time ran out.
That being said, this is symbolic for what the system can do, which is this: it creates an approximation of a concept, rather than an image of the concept itself. Mostly, it doesn't have an understanding of volume, or where up and down is (often you get sections of sky smack in the middle of an image), and it has no means to actually detail the very centerpieces of what you want it to do. As you can see (can you see it?) the castle is extremely blurry und lacking detail. It's an approximation of what a castle might be. That is, despite there being 130 million hits on Google about castles. AI-generated art about other topics (cities, space scapes, etc.) also leaves a lot to be desired.
Your prompts and settings can as detailed and 'clear' as you could imagine, but the result will always just end up as a guesstimate lacking proper depth and understanding of volumes. So: nice toy to play around with, but really nothing threatening or groundbreaking, regardless of what its supporters might claim. Definately not going to replace traditional art (yet), and not worth buying computing time for.
That being said, this is symbolic for what the system can do, which is this: it creates an approximation of a concept, rather than an image of the concept itself. Mostly, it doesn't have an understanding of volume, or where up and down is (often you get sections of sky smack in the middle of an image), and it has no means to actually detail the very centerpieces of what you want it to do. As you can see (can you see it?) the castle is extremely blurry und lacking detail. It's an approximation of what a castle might be. That is, despite there being 130 million hits on Google about castles. AI-generated art about other topics (cities, space scapes, etc.) also leaves a lot to be desired.
Your prompts and settings can as detailed and 'clear' as you could imagine, but the result will always just end up as a guesstimate lacking proper depth and understanding of volumes. So: nice toy to play around with, but really nothing threatening or groundbreaking, regardless of what its supporters might claim. Definately not going to replace traditional art (yet), and not worth buying computing time for.
AI Art Generator: Disco Diffusion vs. Midjourney AI
I have now played around with Disco Diffusion for a whole slew of different images, and tried out Midjourney AI today. As such, I believe I can give a somewhat solid overview of which I'd recommend, and what the pros and cons of each one are.
Disco Diffusion
DD works with Google Collab as a base, needs you to connect it to your Google Drive, and is more technical in its operation, meaning you can see the code that's being used, you can change the parameters therein. That includes image sizes, image iterations, saturation, etc. which is all rather byzantine and needs an additional tutorial document to understand.
Disco Diffusion's great strength is that it's sort of a community project, and that using it is (technically) free of charge. If you're not using a subscription, you can only use it when Google's GPU power is available. That being said, in my time playing around with it you could usually access it everyday, with a few hours of downtime in between.
DD works in batches and steps. Batches means that you create your image prompt and settings, and the GPU will create as many images based on your parameters until the batch is complete (for example, the preset batch number is 50). As for steps, this means it starts with 'noise' and adds more and more detail with each step it calculates. The highest number of steps, IIRC, is 1000 per image. I've been using between 200 and 250 steps.
Keep in mind: it won't do these images simultaneously! You do 1/50 with steps 1-250, then 2/50 with steps 1-250, then 3/50... and so on!
That, however, means that Disco Diffusion is slow. And I mean slooooooow. At 250 steps one image of a batch took me 16-17 minutes to build. One image. Of fifty. Mind, the system will shut down if you stay afk too long, so the best you can get even if you try to keep it running during the night while you go to sleep is maybe 4-6 images.
It is possible that with Collab Pro or other plans you get faster and more GPU time but I don't want to try that just for fun.
Midjourney AI
As far as ease of use and ease of setup go, M.AI certainly is superior to Disco Diffusion. M.AI uses Discord, is intuitive to operate, and generates not one but four images in parallel with each prompt you give. And M.AI is lightning fast.
I'm not kidding.
It does so in less than 15 seconds, and then allows you to refine and upscale potentially all four of them simultaneously. With each step you can further upscale, change lighting, or remaster the chosen images. Again, each step also only takes maybe 30 seconds.
That's maybe 1 minute tops for an upscaled, finished image compared to at least (depending on step numbers) 16 minutes for DD.
Unfortunately, being easy to use and fast comes with the downside of only having a very limited free use contingent. You get 20 prompts for free, and if you want to use it after that you have the choice between a $10 and a $30 subscription model.
That being said, joining the Discord is free, and I'd actually recommend doing so. Browse the different Discord servers and check out the massive number of pretty impressive pieces people have the bot create.
Conclusion
Now, which one is "better"? That's a bit of a subjective take, tbh. Disco Diffusion is free. That's it's really one, big mark that makes it always available. But is it good at what it does?
Ehhhh, sorta, but not really, in comparison. Now, much comes down to personal preferences in style, but I made the experience in my limited time with Midjourney AI that it is significantly better at using your prompts.
Here are two prompts I used for both AI art generators.
cyberpunk city, photorealistic, highly detailed, bladerunner style, tall skyscrapers
Midjourney AI (2 minutes, upscaling included)
Disco Diffusion (16 minutes)
highly detailed Star Wars style spaceship, dark background (same calc. time as first prompt)
Midjourney AI
Now, this isn't really Star Wars-ish, but it definately is a detailed starship that can be used in a variety of ways to support stories etc..
And then there's... Disco Diffusion and this:
I have no idea what in H.R. Giger's name this is supposed to be.
So, my choice after trying both? Midjourney AI! Honestly, I'm really tempted to try out the $10 subscription for Midjourney AI just to try it out for some time more.
Disco Diffusion
DD works with Google Collab as a base, needs you to connect it to your Google Drive, and is more technical in its operation, meaning you can see the code that's being used, you can change the parameters therein. That includes image sizes, image iterations, saturation, etc. which is all rather byzantine and needs an additional tutorial document to understand.
Disco Diffusion's great strength is that it's sort of a community project, and that using it is (technically) free of charge. If you're not using a subscription, you can only use it when Google's GPU power is available. That being said, in my time playing around with it you could usually access it everyday, with a few hours of downtime in between.
DD works in batches and steps. Batches means that you create your image prompt and settings, and the GPU will create as many images based on your parameters until the batch is complete (for example, the preset batch number is 50). As for steps, this means it starts with 'noise' and adds more and more detail with each step it calculates. The highest number of steps, IIRC, is 1000 per image. I've been using between 200 and 250 steps.
Keep in mind: it won't do these images simultaneously! You do 1/50 with steps 1-250, then 2/50 with steps 1-250, then 3/50... and so on!
That, however, means that Disco Diffusion is slow. And I mean slooooooow. At 250 steps one image of a batch took me 16-17 minutes to build. One image. Of fifty. Mind, the system will shut down if you stay afk too long, so the best you can get even if you try to keep it running during the night while you go to sleep is maybe 4-6 images.
It is possible that with Collab Pro or other plans you get faster and more GPU time but I don't want to try that just for fun.
Midjourney AI
As far as ease of use and ease of setup go, M.AI certainly is superior to Disco Diffusion. M.AI uses Discord, is intuitive to operate, and generates not one but four images in parallel with each prompt you give. And M.AI is lightning fast.
I'm not kidding.
It does so in less than 15 seconds, and then allows you to refine and upscale potentially all four of them simultaneously. With each step you can further upscale, change lighting, or remaster the chosen images. Again, each step also only takes maybe 30 seconds.
That's maybe 1 minute tops for an upscaled, finished image compared to at least (depending on step numbers) 16 minutes for DD.
Unfortunately, being easy to use and fast comes with the downside of only having a very limited free use contingent. You get 20 prompts for free, and if you want to use it after that you have the choice between a $10 and a $30 subscription model.
That being said, joining the Discord is free, and I'd actually recommend doing so. Browse the different Discord servers and check out the massive number of pretty impressive pieces people have the bot create.
Conclusion
Now, which one is "better"? That's a bit of a subjective take, tbh. Disco Diffusion is free. That's it's really one, big mark that makes it always available. But is it good at what it does?
Ehhhh, sorta, but not really, in comparison. Now, much comes down to personal preferences in style, but I made the experience in my limited time with Midjourney AI that it is significantly better at using your prompts.
Here are two prompts I used for both AI art generators.
cyberpunk city, photorealistic, highly detailed, bladerunner style, tall skyscrapers
Midjourney AI (2 minutes, upscaling included)
Disco Diffusion (16 minutes)
highly detailed Star Wars style spaceship, dark background (same calc. time as first prompt)
Midjourney AI
Now, this isn't really Star Wars-ish, but it definately is a detailed starship that can be used in a variety of ways to support stories etc..
And then there's... Disco Diffusion and this:
I have no idea what in H.R. Giger's name this is supposed to be.
So, my choice after trying both? Midjourney AI! Honestly, I'm really tempted to try out the $10 subscription for Midjourney AI just to try it out for some time more.
Last edited:
I guess both these AIs (or should we call them VIs) are useful tools, but not this replacement of actual artists. Man, at least these fops got the same scare that your average worker gets when his boss announces they want to modernize the facility.
At least Midjourney seems like a useful tool if you want to quickly sketch out concepts that you then want to base your own artwork on. It's really good, relatively speaking, at what it does. For example it also works rather well with faces and characters. So as a tool to speed up the initial process it's definitely something to consider.
Disco Diffusion however is far too slow to be of any use, and it's ability to create something useful from the prompts you give it is dubious at best.
So, no artist replacement just yet.
Disco Diffusion however is far too slow to be of any use, and it's ability to create something useful from the prompts you give it is dubious at best.
So, no artist replacement just yet.
Bear Ribs
Well-known member
To get good art from an AI you usually have to work iteratively. Midjourney is definitely the one I think is superior at the moment.
I'm not sure how you got only one pic though, normally Midjourney gives you four for your first prompt, then you pick the best one and ask for (V)ariations on it, then repeat. After several iterations usually you can home in on what you wanted, unless it's very esoteric. I never did get it to give me a Pile of Grass Eating a Sandwich I was pleased with.
All the good Midjourney art is typically the result of half a dozen to two dozen Variant runs to remove the dross early images and teach it to produce a good one.
I'm not sure how you got only one pic though, normally Midjourney gives you four for your first prompt, then you pick the best one and ask for (V)ariations on it, then repeat. After several iterations usually you can home in on what you wanted, unless it's very esoteric. I never did get it to give me a Pile of Grass Eating a Sandwich I was pleased with.
All the good Midjourney art is typically the result of half a dozen to two dozen Variant runs to remove the dross early images and teach it to produce a good one.
I got four pics.
Yep, you're right on all counts. You do get the 4 initial images, and you have to play around with them and the prompts a bit. Still, Midjourney's speed and general output far outpace anything Disco Diffusion has to offer.
A few things I'd like to add for better understanding and to emphasize some of the current limitations of Midjourney.
Midjourney, like Disco Diffusion, produces an approximation of what you have in mind. What it puts out will never truly, 100% match what you are looking for, because the AI is simply not smart enough to apply all nuances of your text prompt properly. But, time and again, it might come close.
The aspect ratio that you choose for your prompt is probably just as important as the prompt itself. That is, if you want to create a character and do so as a torso or fullbody shot, it's not really going to work that well with the base image ratio which is a square. Give the system an aspect ratio to work with that emphasizes the dimensions you want to show, like 9:16 or 8:10 (only full numbers, no decimals!).
Midjourney seemingy works rather well with characters, but there are limits: it isn't very adept at displaying more than one character at a time, at least not in detail and to your specifications. Characters also often appear with limited/without limbs, and the system absolutely doesn't manage to have your character hold something in their hands properly.
Lastly, as @Bear Ribs mentioned, the system works iteratively. That is, you will have to create multiple runs at a concept, then take the best of the results, use that as an input, rinse and repeat. The more complex the setup, the more iterations you'll (most likely) need to come closer to what you've envisioned for yourself.
Here's an example of the iterative process when it works really well:
the prompt:
a thin and shy young woman looking like Sansa Stark with auburn locks and tan skin, symmetric face, wearing a futuristic white and dark green business dress with a silver belt with intricate ornaments, artgerm, artstation, fantasy, 4k, detailed, sciencefiction
1). Initial batch of 4 images
2.) 4th image picked to upscale
3.) 4th image remastered (more detail, facial symmetry, cleaner lines); though I did like the ornaments on the previous one better I do not know how to keep them
Midjourney, like Disco Diffusion, produces an approximation of what you have in mind. What it puts out will never truly, 100% match what you are looking for, because the AI is simply not smart enough to apply all nuances of your text prompt properly. But, time and again, it might come close.
The aspect ratio that you choose for your prompt is probably just as important as the prompt itself. That is, if you want to create a character and do so as a torso or fullbody shot, it's not really going to work that well with the base image ratio which is a square. Give the system an aspect ratio to work with that emphasizes the dimensions you want to show, like 9:16 or 8:10 (only full numbers, no decimals!).
Midjourney seemingy works rather well with characters, but there are limits: it isn't very adept at displaying more than one character at a time, at least not in detail and to your specifications. Characters also often appear with limited/without limbs, and the system absolutely doesn't manage to have your character hold something in their hands properly.
Lastly, as @Bear Ribs mentioned, the system works iteratively. That is, you will have to create multiple runs at a concept, then take the best of the results, use that as an input, rinse and repeat. The more complex the setup, the more iterations you'll (most likely) need to come closer to what you've envisioned for yourself.
Here's an example of the iterative process when it works really well:
the prompt:
a thin and shy young woman looking like Sansa Stark with auburn locks and tan skin, symmetric face, wearing a futuristic white and dark green business dress with a silver belt with intricate ornaments, artgerm, artstation, fantasy, 4k, detailed, sciencefiction
1). Initial batch of 4 images
2.) 4th image picked to upscale
3.) 4th image remastered (more detail, facial symmetry, cleaner lines); though I did like the ornaments on the previous one better I do not know how to keep them
Rolling Grasslands (HexMap 36x48)
Rolling Grasslands provides you with a large (36x48 inches) hex battlemap consisting of grasslands with several hills interspersed with patches of forest. The map is explicitly compatible with the official hex sizes and nomenclature of BATTLETECH, but can also be freely used in any other context where you want to play a prolongued battle in varied terrain with plenty of room for cover and maneuvering.
The map size is again 36 x 48 inches and is provided as print-ready (300 dpi) PDFs file as well as high quality JPEGs for your convenience. You get the complete map file, and the map in four parts for easier printing as well.
You can get it at DriveThruRPG for $3.99.