Skip to Content
 

Testing AI art from sketch

59 replies [Last post]
questccg
questccg's picture
Offline
Joined: 04/16/2011
Here's what an attempt at an ISO View looks like!

So I tried to create an "Isometric View" of the Original Rendered Terrain Sample. At first it looks a bit "clumsy" but if you look a bit CLOSER by ZOOMING into the generated image... It didn't do TOO BAD of a job TBH...

Like I said ... It may seem a bit "clunky" at first... But when you ZOOM-IN, you actually can APPRECIATE the HONEST and GOOD JOB the AI did with such an "arbitrary" request on how to "reshape" the landscape.

There are LEVELS and you can clearly see them. Unlike the Original Image that was generated by the AI.

So I guess this needs to be EVALUATED in that ... HOW GOOD? Did the AI do?!?!

Cheers all!

X3M
X3M's picture
Offline
Joined: 10/28/2013
Now that is what I am talking about...somewhere in that picture

You clearly can tell now that one hexagon is higher than another. But not everywhere.

I should not nitpick about some other aspects of the map that got lost in translation. But the elevated stuff with the cliffs is what I am searching for. Although, it kept the rocks for the center piece. But the centre should be the same design as on the left and right that has forest on them. And only apply rubble when rocks are in the centre portion.

I have seen some of these before on the internet as well. Somehow, the AI seems to be losing track of some hexagons. Here I see it slowly happening as well. I guess it is because it is not given the highest priority.

Making a map like this. And including the game mechanics to depend on specific maps properties. Each aspect has 100% priority. But it doesn't have to be. Perhaps the AI can make some maps. And then we see what rules can still be applied.

Or...I should try out another way. Such that hexagons can be stand alone. And fit together seamless. Thus, removing the strips around each hexagon when designing maps. And consider each hexagon as a standalone that way.

Mission facts; terrain features that got lost:
Dessert; slows down certain units)
Swamp; allows for ground units movement, you clearly cannot tell anymore)
Some parts have become invisible to the eye since it is very isometric now.
It is very hard to tell the actual height difference. How much do we go up or down? In Heroscape you can know. But not here.
How much room is there in a forest?

***

Amazing job done by the AI.
Just not exactly what I am looking for....yet.

Can I make compromises with the given map?
I need to work my ways towards it.
I think I should not keep looping in this circle.
And find some time to draw some proper cliffs.

Isometric. But visible.

larienna
larienna's picture
Offline
Joined: 07/28/2008
Lately, chatGPT has been

Lately, chatGPT has been starting to block me even with just text discussion.

I don't think you train your own AI with stable diffusion, you give reference images. Training would take much more time, data and processing power. When we did it in one of my class, we had to use google colab, and the free plan was not enough for me.

If the model is not as cool as other AI, it should still be better when trying to do it by hand even if I had the skill.

Your maps are amazing, they are more awesome than the battletech maps I used to have. Still, as a board game, you need to be able to identify which terrain is where to apply the rules, for video games, its less and issue.

For my personal use, I would preer the map to be modular, which means have individual hex textures that I could spam around to allow computer generated maps in game. Having more than 1 texture for the same terrain could increase the variety. This is why the "Wizard king" map design was very interesting, because I can have a logical map and a display map where I plant the icon where I want. Especially for maps where hex edges have a different terrain than the center.

You could have maps a la starcraft with low and high areas with ramp between them. The 3D map is interesting, but you might need to all the elevation level in the hex like in battle tech maps. My only complaint here, and I am cherry picking, is that some of the hexes moves out of the ground like if there was a chasm. Its not a smooth transitions. Still it reminds me of Inkscape.

questccg
questccg's picture
Offline
Joined: 04/16/2011
You CAN train Stable Diffusion

larienna wrote:
...I don't think you train your own AI with stable diffusion, you give reference images...

Just Google "Can I train Stable Diffusion" and you will see Google will tell you YES you can train the AI and produce a file that can be later used to produce content similar to your Art Style.

Quote:
LoRA (Low-Rank Adaptation): Most users on Reddit agree that this is the best and most popular method because it creates a small file (tens of megabytes) using just 4 to 20 images.

I know this because an artist wanted to see if she could get the AI to help make commissions due to a project having a high demand on "volume" (High Quantity of Images/pieces of art...) Something in the volume of 150 to 200 pieces of custom art.

But realize that "Stable Diffusion" is not as perfected as OpenAI's ChatGPT. Yet it could be interesting to SEE if you CAN Train and AI to do ART in a specific way or not... IDK. Like I said that artist said she was looking into it. Not sure how far she got!?

In any case... You are not an artist. So training is pointless to you. Since you own NO ART. I was just saying it is POSSIBLE because I had discussed it with her and another Game Designer who was using it to produce Comic Book Art.

So... Now you know it is possible. How much input data do you need??? IDK. You can of course "Google" some more and read some Reddit forum posts about the subject of training "Stable Diffusion"...

Cheers.

questccg
questccg's picture
Offline
Joined: 04/16/2011
As a side note...

The artist in question was being asked to draw "Comicbook-esque" Dragons. And she has her own style of illustration and thought it could save her time when creating characters.

Basically she wanted to input a PROMPT and get some kind of unique illustration given that prompt and HER "Art".

Don't know what happened to that project. I did not follow-up in the discussions because she "banned" me from chatting with her after we had an issue with some incorrect assumptions concerning various printers (the shops not the machinery).

Some people just overreact to everything and everyone. I guess she wasn't having a good day and my post didn't land as being 100% accurate and she decided to ban me from messaging her.

Whatever... I don't need to message her. I was just curious as to her "project" to train AI and how it worked out (or not)! But since I am banned I cannot message her anymore. Short fuse... Impulse reaction and voila... Someone that I cannot message and ask how training went.

Like I said... Some people!

Note #1: I guess she didn't want all her dragons to look like "Charizard"... Hehehe.

questccg
questccg's picture
Offline
Joined: 04/16/2011
Here's a Link/URL with some Stable Diffusion Art

https://www.bgdf.com/forum/game-creation/new-game-ideas/epic-battle-pnp?...

That's what some of the "Cartoon"-like characters came to look like. I don't know how the prompts work in "Stable Diffusion"... I haven't heard from @FrankM for quite some while now (like 2-Years)... He was a regular for a decent amount of time... Like a year or two... But like I said, I haven't seen him on for a while.

You can scroll through the thread for different views of his "Comic book"-esque "Epic Battle"... I also posted an improved layout for his Sample Page...

It's not as realistic as ChatGPT... But if you want "Comic Book" type art. It ain't that bad either...

Note #1: Here is @FrankM account... In the event you may want to PM him:

https://www.bgdf.com/users/frankm

questccg
questccg's picture
Offline
Joined: 04/16/2011
More about Training the AI.

questccg wrote:
LoRA (Low-Rank Adaptation)...

Googling some more and it says you can have hundreds of images... And recommends you use different "backgrounds" to help the AI translate that as ways to vary the background on PROMPTS.

Google wrote:
Variation: Change the background, clothing, and camera angles across your photos so the AI learns the core subject rather than a single repeating background.

Or this tid-bit on CAPTIONS:

Google wrote:
Good Captions: Write accurate text descriptions for each image to help the model learn faster and more accurately.

I made a mistake in my Google Search and it came up with more information ABOUT how(?) to train Stable Diffusion BETTER.

My mistake was:

what is the more images you can use to train stable diffusion

It produced a better response than "the MOST images" which returns a response saying you can train an infinite amount of images... Sometimes we make mistakes too... And the result is something BETTER than we had anticipated!

Cheers all.

X3M
X3M's picture
Offline
Joined: 10/28/2013
I still think

That I need to see if I can do without those in between things.
Instead, hexagons as standalone. And then have the higher ground bleed into the lower ground. That would make much more sense.

Imagine from a 3D print. You got this hexagon that exists. But has to go to a lower ground in the next one? If you want a nice little ramp instead of a drop. You have to add the ramp and place it on the outside of that hexagon. It doesn't have to be big. Just enough to indicate it is a ramp, not a cliff. And the lower elevation will remain flat as well. I think, this is a good way to think in 3D. And thus, I should find a way to make this happen in my 2D maps. Before returning to AI.

AI seems to be.... well, form time to time, you don't even know what it is doing. A very stric guidance is needed. And often, things are even forgotten. So, you need to keep reminding the AI in how to do things. Have some tests ready to see if the AI still does the job.

This goes for calculations (just discovered that the AI cannot calculate the numbers of a hexagon). No doubt, this goes for art as well.

questccg
questccg's picture
Offline
Joined: 04/16/2011
Went back to the Original SOURCE file and...

I did it in ONE pass... To both convert it to ISO and to RENDER the scene too:

Like I said, I think this VERSION is the BEST one YET! Which means with the right kind of SOURCE Map File... And the right prompt (and experience), I can bet you that you'll get this QUALITY.

It's NOT 100% as the Top-most path got LOST... But it's about 90% and even better in terms of LEVELS and CLIFFs too!

Feel free to comment and tell me what you all think!

Best.

Note #1: I may propose that the mission STARTS at the TOP-MOST Hex (in the middle) and if you go LEFT... You reach an impassable hill that drops... Which means you need to go RIGHT and follow the ENTIRE trail until you make it to the TOP-LEFT Exit to the Level... Could be cool, no???

Note #2: And might I add... the TOP-LEFT path is also "blocked off" due to the height. So maybe if you reach that Dead-End ... You'll meet up with some more TROOPS (Missile Launchers) who parachute in to help return back down to the ONLY viable exit to the LEFT which has a bunch of troops including Tanks ... Which your newly acquired Missile Launchers are available to help finish of the "Mission"... Something like that.

Damn it looks good. Too bad for those two (2) inconsistent PATHS at the TOP-LEFT... But like I said, sometimes 90% of something GREAT is better than 100% of something sucky.

Hope you LIKE it... At least a LITTLE BIT! Hahaha.

Note #3: And yeah the "Sand" at the top looks more like Savannah than sand. But at least this time, it actually did something GOOD with the colors of the "grass". They're intact and they clearly blend in from grass to savannah ... Not so bad afterall...

X3M
X3M's picture
Offline
Joined: 10/28/2013
I was like

Hey! It got the middle right, right.
Then I proceeded to look top left...the road!
It is also almost gone a bit ahead.
The swamp is a bit of nitpicking at this point.
But yeah, a good job indeed.
Especially that the savannah has returned.

There is one thing that all renders got wrong. It is also very hard to miss. Let's see if you can guess what it is :D

I rather see that you save your points. And use it for something personal. And me and my busy life kinda would like to look into the 3D prints and 2D sticker pasting. Whereas the 2D sticker pasting would be more like that very first rendering the AI did.

You know, what I described in Discord? Maybe I should draw some schematics?

questccg
questccg's picture
Offline
Joined: 04/16/2011
I have no clue TBH!

X3M wrote:
There is one thing that all renders got wrong. It is also very hard to miss. Let's see if you can guess what it is :D

No freaken clue!? Feel free to share...

larienna
larienna's picture
Offline
Joined: 07/28/2008
Good to know that you can

Good to know that you can train your own AI. But I suspect you need thousands of images to do the training.

I trained a neural network to distinguish dolphins from sharks in one of my class. I don't remember how many pictures I had, but its easily above 1000.

So even a prolific artist could not make its own art bank and train from it. Still, it would be cool if you could do it. Or generate something to use a base artist can work from, or use it to enhance a drawing like I did, but maybe at a different step.

We will see how things evolves.

questccg
questccg's picture
Offline
Joined: 04/16/2011
Here's what I mean

larienna wrote:
Good to know that you can train your own AI. But I suspect you need thousands of images to do the training...

That may be true but one thing for certain is that ChatGPT "learns" from your interaction with the platform. Like I had generated some tiles for @X3M a few month ago... And it remembered the style and feel when it made the ISO View...

So it ADAPTS to what you ask of it and the results you get back (satisfactory or unsatisfactory...)

Just my observation ... Because whenever I mention "G.I. Joe" it thinks about my "Mystery Card Game" and usually provides (sometimes) the wrong SOURCE or too much input from previous prompts.

In any event... I'm positive it adapts to how YOU use it!

Cheers.

larienna
larienna's picture
Offline
Joined: 07/28/2008
Using reference information,

Using reference information, or information from the memory is not the same thing as training a model. The reference information is more like a parameter to the model.

questccg
questccg's picture
Offline
Joined: 04/16/2011
Is this like Short-Term Memory vs. Long-Term Memory???

When you say "Reference Information" is that like "Short-Term Memory"??? I mean the AI makes up a CONTEXT in which prior conversation had occurred and "sorta" uses this to figure out what it should do given a "Text-based Prompt"??

And then Long-Term Memory is like content that was used to TRAIN a Model? Like it normally would do it as "X" as a method... But now we input "A-B-C-D-E..." as Inputs or parameters to the Model?

Something like that??? Or am I completely lost?? Please let me know. I am more than fascinated by AI. It's a huge leap in computing TBH. They probably never know they would be so successful in programing the AI. And look at it now?!

Cheers.

questccg
questccg's picture
Offline
Joined: 04/16/2011
Here is the 98% version...

I did it in ONE LAST attempt... To both convert it to ISO and to RENDER the scene too and try to restore the top-most road:

Like I said, I think this VERSION is the BEST one YET! Which means with the right kind of SOURCE Map File... And the right prompt (and experience), I can bet you that you'll get this QUALITY.

It's 98% because ATM there are no RAMPS to speak of and that could be a difference in terms of the possible types of compatible terrain given that the AI "could" understand RAMPS.

This version required ONE (1) MANUAL "tweak" ... It did require ME going into PaintShop Pro and making one manual addition.

Feel free to comment and tell me what you all think!

Cheers all.

Note #1: The reason I say it is 98% is because of the following FACTS:

  1. The colors match the SOURCE Map File 100%.

  2. The Savannah on the Top-Most road is correctly in a Yellow-ish grass.

  3. The Road in the TOP-LEFT is pretty well done as the Hex Tiles are more-or-less even.

  4. The Road in the TOP-RIGHT was added MANUALLY (Using PaintShop Pro) and tweaked a bit to blend in better.

  5. All the other elevations and CLIFFs are respected.

That having been said, the water has lost some of it's luster ... But that's not a major focal point of the design.

So all-in-all not too bad especially when done in ONE PASS!

X3M
X3M's picture
Offline
Joined: 10/28/2013
Another thing to try

I posted details in discord:

- Remind the ai to use the correct textures.
- Inform the ai to have the correct art for high to low ground, thus the center hexagon should not have rocks going up first... but go down immediately.
- Remove the center dots.
- Keep the hexagons, obviously.
- Add the numbers (Use the one with the numbers). Centrialize them?

questccg
questccg's picture
Offline
Joined: 04/16/2011
No miracles...!

X3M wrote:
- Remind the AI to use the correct textures.

You can't REMIND the AI of JACK SH!T. It does as the SOURCE MAP given directions and produces an output based on the ACCURACY of the PROMPT. The more GENERAL the prompt ... The MORE likely the results will look accurate.

X3M wrote:
- Inform the AI to have the correct art for high to low ground, thus the center hexagon should not have rocks going up first... but go down immediately.

As I say... You don't have GRANULAR input into the AI to make it decide this. It does as it sees fit based on the SOURCE image. It sees rocks so it made the rocks into rocky-hills. Do you understand now???

X3M wrote:
- Remove the center dots.

Just REMOVE them from the SOURCE image.

X3M wrote:
- Keep the hexagons, obviously.

It does as it sees fit. Some HEXES are intact ... Others may look a bit deformed but that could also be the ISOMETRIC Angle too...

X3M wrote:
- Add the numbers (Use the one with the numbers). Centralize them?

Use a Paint Program with LAYERS and add a layer(s) for Vector Text. Don't expect the AI to do it.

And with all these RULES... Do you think the AI will render better the MAP??? Too much prompting results in more CRAPPY OUTPUT. Which again is something that I have mentioned before. If there are things you can DO AFTER the MAP is generated ... Like Add Numbers and so forth... You do it in your Paint Program not via the AI!?!?!

You seem to expect that AI will be a single pass and solve all your issues. AI is the terrain and hexes, you want NUMBERS add a Vector Layer and add you numbers on top of the JPG or PNG file. Don't expect the AI to handle it... You will be sadly mistaken how POOR the AI reacts to "artifacts" which are not necessary to the TASK AT HAND: which is rendering the terrain.

Don't be mad, I'm trying to explain because you seem to think that the AI is a miracle tool that does it all. Remember the FINAL MAP I gave you was EDITED in PaintShop Pro. Why? Because it forgot to make the ROAD EXIT on the RIGHT-HAND SIDE... I put it in manually using a Paint Program.

So even the IMAGE that you HAVE was EDITED and FIXED because the AI FORGOT a detail and removed the EXIT...

Anyhow... Please don't be upset or anything. I'm just trying to MANAGE YOUR EXPECTATIONS and AI is not the easiest thing to deal with in regards to what it produces from a SOURCE image.

Sincerely.

Note #1: Think about AI in this way:

Quote:
AI is a temperamental ARTIST.

Some times he uses the colors HE feels like using and sometimes HE takes liberties to draw what HE sees fit and not at all exact as an INPUT file is given.

He does what he wants... Sometimes you can correct him but then again he'll change details you LIKE and do something ELSE that you didn't want like make your SWAMP more like LAKES or PONDS because he does SH!T to annoy you... Get it?

He thinks he's smarter than the HUMAN asking him to PRODUCE ART... And so often does what you DON'T WANT or changes TOO MUCH or NOT CORRECTLY, etc. etc.

That's what I am TRYING to explain to you... But instead you are putting more rules and such that make it more and more complex and he will do as HE WANTS not YOU!

questccg
questccg's picture
Offline
Joined: 04/16/2011
Here's what I mean

I figured because you don't understand how the AI works... I'll show you some FAILED attempts at TRYING to fix sh!t. Here you go:

This was an ATTEMPT to restore the "Savannah" to the TOP of the MAP... See what the AI did in terms of MISTAKES??? There are a bunch of them... I won't tell you all of them ... But you can SEE it can be REALLY FRUSTRATING every time YOU try to ask it to FIX something it does it HIS WAY!

A lot of things wrong with this EDIT VERSION.

That's what happens when you TRY to FIX things you don't get in ONE PASS and you rely on "intermediary" MAPS to try to FORCE it to do what WE WANT! But no as a FRUSTRATING ARTIST ... He does as he sees fit and again this is one of three (3) image files I can generate per day.

I'm not bullsh!ting you... I'm telling you as the GUY who worked with the AI to TRY to get you the BEST possible OUTCOME! AI can become VERY frustrating. And you seem to think that IT CAN DO EVERYTHING. And I'm trying to explain: "Do as much as you can to ensure it gives you the RESULT YOU WANT!" And then add sh!t in a Paint Program which is NOT related to the AI Generation...

Do you see my point yet???

Note #1: I have OTHER "intermediary" images that I can show you where the PROMPT was meant to correct the Savannah ... And then the Desert Road DOES NOT connect with the Grassy Road. Another wasted generation and again the AI doing what HE sees fit.

I can show you that file too... But again ... It's more wasted generation and time taken and frustrating results because the AI "is a temperamental ARTIST"! You can't believe how HE makes so many mistakes and you are like: "WTF...??? Where(???) did that come from. I didn't ASK for that!!!" And yeah ... It does GARBAGE even if you think your prompt is SMART. In many cases HE does what HE WANTS and you've got to HOPE the result is REASONABLY GOOD. It may require a Photoshop Here-and-there ... Nothing a Graphic Artist cannot handle.

questccg
questccg's picture
Offline
Joined: 04/16/2011
Here's the BEST Version to be used with 3D...

Well I think the ISO Views are too "ugly". The may be more CORRECT, but they produce too dark terrain effects which also seem like "chasms" instead of "Cliffs" (Holes in the ground).

Here's the LATEST and GREATEST version so far:

This attempt REQUIRED a "manual edit" seeing as the ROAD on the Right-Hand Side disappeared and was missing, I had to add it in manually in Paintshop Pro. But rest assured it looks seamless and there was some smoothing going on in that area to ensure that the road Exit/Entrance appears reasonably OK looking.

I know @X3M is working on 3D Tiles and he seems to FAVOR this version over the ISO Version because I think everything just looked a LOT DARKER and even if in terms of application was more "realistic" ... This version of the REGULAR View is much better TBH (IMHO!)

We'll see what @X3M thinks of this version. I for one think it's the best NON-ISO Version so far!

Sincerely.

questccg
questccg's picture
Offline
Joined: 04/16/2011
So I made another test tonight and...

It turns out that the AI "cannot" comprehend the concept of "RAMPS". I could easily visualize it... But the AI CAN'T. I tried a sample with three (3) RAMPS to see what it would produce and how clean it would look. Well I'm here to tell you that IT DID NOT WORK.

The best looks like a RAMP that doesn't quite meet the NEXT level. The other two (2) looks like some dirt just put there and for no logical reason without any real attempt at even LOOKING like RAMPS.

I'm a bit disappointed... But I might try something else... IDK What maybe some kind of ROAD could help render the RAMP... IDK. It's going to take more DETAIL and some other type of Graphic Design to make it happen.

I will continue to work on Attempt #2 ... As Attempt #1 was a COMPLETE and TOTAL UTTER FAILURE.

Sorry to disappoint. It's rare that I post "NEGATIVE" content on BGDF.com... But this time, I had no choice: the end result was totally unrealistic.

Sorry all.

Update #1: So I added a dirt road with a rocky texture. It did NOT work... If creates a sort of HILL on a HILL... And now at all a RAMP with a road. I'm all out of credits I will retry tomorrow with a different texture... We shall see... TBD.

larienna
larienna's picture
Offline
Joined: 07/28/2008
So doing a 2 level terrain

So doing a 2 level terrain like starcraft where you have ramps to transition from one level to another is complicated to explain to AI?

X3M
X3M's picture
Offline
Joined: 10/28/2013
Yip

Starcraft has clear art for showing what cliffs are. And ramps are only north-high, bottom-low.
The art used in the new maps these days got the other types too, but that looks just awefull.

Starcraft 2 is 3D. And viewed from an angle.

I tried to see if the AI could do something similar to C&C. But that is just not done either.

So, back to basics I guess. A top down map, with 2D art. And then simply the numbers of height on the hexagons.
Once I get to 3D printing with 2D stickers. I would simply expand the stickers in a certain way, in order to fit on the ramps. As for cliffs in the game, viable when the height is "6" or more. Then again, 1 set of ramps that is just 1 high is good enough.

This:
https://img.itch.zone/aW1hZ2UvMjY2ODI5MC8xNTkwNTYwMS5qcGc=/original/pYWs...
Seems to be interesting to work with.
Here the hexagons ARE, the ramps.
But still, nothing for the back of a hexagon?

Then we have this:
https://lh5.googleusercontent.com/-pIMgk74fQGg/UCevCkXHGNI/AAAAAAAAAVg/Q...
Even though, the grid are squares. You can see that one ramp goes down when going to a north direction. The slope just has to be not toooo steep.

Then there is the orientation of hexagons. Perhaps the "flat" ones are not good. And you need to use those that "stand up. Such that you have 2 north sides (west/east), a west and an east and 2 south sides (west/east).
The "flat" has 3 north (west/north/east) and 3 south (west/south/east) sides.

The AI seems to be able to think of something with the 3 south sides. And for standing with the 2 south sides.

But the north sides are non done. And I didn't even look for the west and east sides in the standing version. I guess, there is hardly anything done in this regard. Because the AI needs to learn from examples.

Isometric will do me no good if I make maps that have an extreme drop in heights. Unless you are always looking at a mountain in the background.

Like in Snake Rattle'n'Roll
The entire map:
https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcSGN08zYihv-OAMgUer...
No ramps, but still forced to go up when going north.

questccg
questccg's picture
Offline
Joined: 04/16/2011
As @X3M said...

larienna wrote:
So doing a 2 level terrain like starcraft where you have ramps to transition from one level to another is complicated to explain to AI?

Yip... I tried with a rocky texture and a path ... It turned out sh!tty too... It produced a terrible result that looked like a Hill on a Hill. UGLY!

So I re-did the road more like the original map you see in this thread. I will see how the AI processes a road with THAT "texture". But I have no credits until 9:00 PM EST tonight. That road surface is NEAR-IDENTICAL to the road with the fully rendered Terrain in this thread.

I tried some tricks like make the RAMPS "purple" hue and it did SOMETHING but that wasn't good enough. I showed @X3M what it made and clearly it's NOT good enough TBH. It's poor with brown textures unless they are on grass (Green). IDK why... I guess that's how it was trained!?!?

It kinda gets a BIT what a RAMP is... But not like in "Starcraft 1" which were OBVIOUS and you could travel to the next level of terrain via the ramps. AI so far... Doesn't seem to understand that... I'll U/L the crap and show you... In the next comment...

questccg
questccg's picture
Offline
Joined: 04/16/2011
Here's the CRAP AI made...

In this version, what I tried is to mark the RAMPS as being a "Purple Hue" and I asked to render this as a "Ramp".

Here's the CRAP it produced:

So the TOP-MOST Hex looks like "some-sort-of RAMP" that starts but never reaches the TOP LEVEL of the terrain... Bloody awful...

The two other hexes produced hexes with some kind of "interference" of Rocks in from of it. Which is WEIRD because there are NO ROCKS in FRONT of the RAMPS. Why(?) the AI did this??? IDK... But it's ARTIFACTS that I DID NOT ASK FOR!!!

This was using a sand texture and a brown rock texture. Without ISO View, this is a normal render, IT SOMEHOW UNDERSTOOD 2 Levels of Elevation ... But could NOT make the RAMPS connect. Just weird.


Tonight I will make the last couple attempts. What I did was change the texture from Rocky-Hill Texture to a Brown Noise Texture exact like in the COOL map with the road and swamp/lake...

We'll see what it produces... I will report back later after the attempts because I want to see WHAT it can produce and maybe there are other OPTIONS too... TBD. I am not 100% sure yet.

The AI is very unpredictable so we'll all have to wait and see...

Cheers.

questccg
questccg's picture
Offline
Joined: 04/16/2011
What about ROADS???

Well I was thinking what if I made some roads with the same texture as our "pretty map/terrain" with the Road, Swamp and Ponds.

Here's the NOT-SO-BAD result it produced:

So I am a bit "miffed" at this result. Clearly it seems to MAKE a "road" that can go up to a higher level... But it still looks a bit FORCED rather than appearing to be 100% natural.

I can say for CERTAIN ... This type of "Scenario" is FAR FROM IDEAL TBH. I was picturing more larger segments of terrain to occupy more of the SIDE it is linked to. But this puts a bit of a burden on LAYERED Maps with several levels...

You basically NEED to put in a ROAD if you want it to look "somewhat" OK. I say "somewhat" because if you've played "Starcraft 1" you'll know EXACTLY the TYPE of RAMPS that I was aiming for.

But I guess this is kinda OK. Not too sure how I feel about this version TBH.

If anyone has an opinion, wants to add to this some feedback, or have any type of question feel free to ask away and post up your response.

Cheers all.

Note #1: So this version looks very "Painted on". What do I mean? It looks like someone took a BRUSH and just painted the ROAD atop the terrain with BROWN PAINT. Yes it's better than the previous attempt... But in terms of "quality" ... It looks VERY disappointing.

You can't believe how SENSITIVE the AI is to "Textures". The wrong texture yields the WRONG RESULT. Seriously I attempted the ROAD with a "Rocky Texture" and it made an even HORRIBLE result.

At least with THIS ROAD TEXTURE ... I looks like a ROAD... Even though it is NOT 100% perfect.

Note #2: I wanted to add that THESE "Roads" add more realistic detail to the path. What I means is that the road has rocky paths on each side of the Dirt Road and it looks pretty realistic. Where it FAILS a bit is the "3D or Elevation" effect it does because it doesn't look all that great TBH.

I will re-try tomorrow with a WIDER ROAD... It's just another attempt to see HOW the AI performs with a variation on THIS "version". Tomorrow... Tonight no more credits left over...

questccg
questccg's picture
Offline
Joined: 04/16/2011
I made another attempt and well...

The result is pretty decent TBH. I didn't produce any images... It wasn't a complete Map... Just certain Terrain Tiles that I wanted to "test out".

And I have to admit the end result looks not too bad. When I say "not too bad" what I mean is that there MUST be a "Road" for RAMPS to exist, the road needs to be long enough for the AI to recognize that it's a ROAD and that if it's going from LOWER Ground to HIGHER Ground, it needs to "sorta" look like a RAMP.

I put in the "Hex Lines" and that made the Road look MORE REALISTIC because it kinda broken the road up and made it FEEL and SEEM more consistent with the rest of the Tiles that I tested.

So the 4th Version looks "decent".

It's NOT like "Starcraft 1" which had amazing looking ramps up to higher elevation. In this AI Version, it looks like ROADS going up to higher terrain level ... Not too bad. But again nothing like the AMAZING RAMPS in "Starcraft 1". And I mention the original game because it was ISOMETRIC and simulated 3D. "Starcraft 2" is fully 3D and just can adjust your viewing angle.

In any event... I like the end result of this 4th Attempt. It looks MUCH better than the previous attempts.

Regards.

X3M
X3M's picture
Offline
Joined: 10/28/2013
Personally...

I think we should abandon the idea for isometric for now.
It doesn't suit my game mechanics. Even though some renderings where beautifull enough to be used by others.

You could still try to get ramps and cliffs to be noticable in an isometric map. But I think that if we want to make real progress. We should go back to the basics. See if the AI can "sketch" singular hexagons.

Just to see if the AI prefers the entire hexagon to be a type of terrain. Or that the edges of said hexagons can be made that way.

I see a lot of examples out there that all have the hexagon itself having the property of being a cliff or ramp or just flat terrain. And perhaps that would be for the best for graphics. Since the entire hexagon can be used for indicating how high the terrain difference will be. The down side is that you need to make much more designs. Another downside is that the hexagon is cut into 2 pieces. And you need some sort of indication for the unit(s) standing on low or high ground.

Still, isometric...? I guess that even those hexagons, will be super difficult for the AI to work with.

I cannot find good stuff on the internet. We got this crap though:
https://www.reddit.com/media?url=https%3A%2F%2Fi.redd.it%2Fu073u7uwm3o11...
Someone asking the questions...not sure if they got the right answers.
It seems that isometric has more problems than just the graphics.

I also noticed that most SC2 maps have most ramps viewed as if you are playing SC1. Even though you look at a 3d map. Why? IDK.... maybe people that design the maps have problems with it visualizing it for themselves. Hence the option to have symetry in maps came along somewhere....

https://starcraft.fandom.com/wiki/Acropolis_(StarCraft_II)

Here you see shadows working for you. But still, the cliffs on the backside are hard to see anyway.
And it is only 1 level at a time when applying ramps.
If....the AI manages to have proper visual ramps and cliffs. And also has a way to show a double or triple height for the cliff. I might consider changing the rules a bit to meet in a middle ground.

Do note.... you are looking at a 3D map in that last link. You can clearly tell that the north is more isometric than south.

https://i.imgur.com/gLkRoV0.jpeg
This one shows that a height of 2 simply gives the same effect on the back.... if there is room for units hiding? Yes, most RTS games make use of this fact. And that makes things even more complicated.

In C&C, we had infantry units hiding for the commando in mission 6.
In SC1, we even got an extra roll on accuracy for hiding behind objects. Or standing on high ground. And I got rules for that too. But we cannot have a player place soldiers behind the board.

Hence, a 3D printed board is a solution for most problems. So, if the AI can make art that is suitable for a hexagon. And make art that is suitable for cliffs viewed from the side. Like, really a sideways viewpoint. Then also, make art for the bleedouts, in other words, the ramps. Then we might be getting somewhere. And we can place units on those hexagons. If there is a forest, well, one or 2, 3D printed trees can easily be added, as long as they can make room for units that need a spot as well.

So, again, let's abandon isometric for now. I must whip up in UTed2 THE 5 blocks. Make some examples. And we work from there I guess.

The plan for now:
- UTed2 for the map initial plans.
- 3D printer making the blocks as well.
- 2D printer making the stickers with the art to paste on the 3D printed blocks.

questccg
questccg's picture
Offline
Joined: 04/16/2011
Do what you have to do...

I've (personally) and don't be offended... have had enough fighting the AI to do the RIGHT thing. It sucks sh!t and behaves like a "Frustrated Artist" that only want to ADD sh!t and do sh!t in it's own way and takes liberties to remove some rocks or remove an exit, etc. etc. I'm tired battling with it.

Work on your tiles... And IF you come out needing some kind of STICKERS, let me know. But FOR NOW... I am DONE with that cr@p!

I can't believe how STUPID it can be in adding sh!t I NEVER ASKED FOR!!!

If you need specific STICKERS, I can probably make them MYSELF. But you are the Captain and it's your PROJECT. I just wanted to see what can be done. I've seen enough. I've played around with the AI enough too. I know what IT CAN DO and WHAT IT REFUSES TO DO.

So if you want to share progress with the 3D Game Tiles... Feel free to do so. IF you have something specific in mind, let me know. But I personally will take a step BACK because it's pointless for me. I don't need Game Tiles... And if I did, I would handle it a bit differently.

I too would design it ONE HEX at a time. Figuring out how many HEXES I NEED. And then go for ONE THEME with like 20 to 25 Game Tiles. And then build like 6 to 8 Missions using those SAME Tiles to boost the VALUE of the product: see we give you 6 to 8 MAPS you can PLAY. And of course you are welcome to design your own too... But I'm NOT a "Wargamer" and I don't particularly LIKE "Miniatures" and I'm not into painting them either, etc. etc. If you want to do it as a HOBBY FINE... But I don't do this for FUN, I ultimately am trying to market products that people WANT to buy.

That concludes my personal opinion (IMHO). And whatever you all choose to do is up to you, enjoy the process and hopefully you'll all have products that you enjoy and are proud to "show-off".

Respectfully.

Syndicate content


forum | by Dr. Radut