Others No Trace Of Luck [Pre-Release] [Development Thread] [Godot]

[NoTraceOfLuck]

Hi all, I am working on a new game. It's built in Godot. Full disclosure up front: All art is AI generated. But I put a ton of work into generating and selecting great art. I have generated 10,000+ images and have only use less than 0.001% of those in the game, and most of the images require significant manual work to ensure consistency and quality.

I will use this thread to post some updates for the 2 people out there who might care while I work on it. This content is F95 exclusive, I have no patreon, am not accepting any kind of donations, and am posting my content exclusively here for free.

LATEST UPDATE: 11/6/2024

Spoiler: Latest Status Update

You don't have permission to view the spoiler content. Log in or register now.

Spoiler: What I'm Working On Next

You don't have permission to view the spoiler content. Log in or register now.

Spoiler: Known Bugs

You don't have permission to view the spoiler content. Log in or register now.

Spoiler: Longer Term Goals

You don't have permission to view the spoiler content. Log in or register now.

Controls: spacebar to skip dialogue quickly, H to hide dialogue, esc or right click for menu

Get the latest version here:

You must be registered to see the links

Screenshots

 

[NoTraceOfLuck]

New version just uploaded. I have the poker running really solid now. Add character art to the conversations, and built support for short video animations (AI generated). You'll see one at the very end.

New version download link:

You must be registered to see the links

Image:

 

[NoTraceOfLuck]

New version uploaded. From a content standpoint, only minor changes. But in the back-end there have been some serious overhauls that will allow me to move much faster.

Changes

• Added support for fade-in/fade-out transitions, selective pausing / disabling of the UI to make certain scene transitions (especially video transitions) look a lot better.
• I removed that 1 character cg that I mentioned felt out of place
• I fixed the issue where mashing Spacebar could cause you to inadvertently make wrong choices
• (Not user facing) I modified the open-source scene management tool that I use to better suit my workflow, allowing me to configure backgrounds, character art, and dialogue all in one screen. This will allow me to build new content much faster.

Here's 2 screenshots, one is just for fun (it was already in the game before) the other is a screenshot of my updated scene tools (notice I can now select a Background and Character State on each dialogue node. I could not do this before

Latest download:

You must be registered to see the links

 

[NoTraceOfLuck]

New Phone Who Dis??

I spent most of my time making the phone messaging screen really solid. I'm very happy with it now. See gif below (note it appears laggy in the gif, but in the actual game it is buttery smooth)

Note the rest of the phone still extremely early stage and WIP:

• The only conversation that works is Lisa. The 'Poker Girls' conversation will crash the game
• The screen where you select the conversation is very early stage and ugly
• The other app on the phone homescreen is non-functional right now (it will be a change wallpaper app)

The phone will play a big part in this game, so I'm happy to get it working so smoothly here.

Coming up next:

• Clean up the rest of the phone. The wallpaper changer app will likely come later, not in the next update.
• There is a distinct lack of sounds effects right now, I'll work on getting those added too
• Trying to get some animations in various places to make the game feel more alive. (Mainly character art transitions & phone transitions)

Download if you wanna play:

You must be registered to see the links

 

[NoTraceOfLuck]

Big update today, spent my whole weekend on it:

Changes

• Made many visual changes to the phone, much less ugly now
• Made Wallpaper app functional + wrote all underlying functionality to enable wallpaper unlocks at arbitrary locations throughout the game. Currently there is only 1 unlock, at the end of the intro
• Added unlock notifications + framework to make any arbitrary notification
• Fixed: Bug where dialogue box would lose focus occasionally, it is now much easier to spacebar mash through the dialogue
• Fixed: Bug where dialogue box continue button would sometimes never appear
• Fixed: The fullscreen experience is now better. This game intentionally uses an abnormal aspect ratio (7:9) and so it will never fullscreen well, but you now get a better experience in fullscreen than before

Coming up next

• It is important than I keep a handle on the memory usage of this game. As much as I would love to load every single image in the entire game up front, that will cause lag as the game gets larger. Today it is not a problem because the game is small. But as I add more and more content and images, it is important that the unused images are unloaded from RAM. My next focus is ensuring I have all this squared away so updates can continue quickly in the future
• 2nd phone conversation proof of concept
• Audio pass - There is almost no sound effects or music in the game. I'll work on creating some and getting them added

Download:

You must be registered to see the links

Image 1: New notification system

Image 2: New wallpaper app

 

[NoTraceOfLuck]

Holy fuck that sucked.

Latest changes:

• I completely restructured the game files. And holy fuck did it suck. I am never doing that again. I don't care how unorganized it gets, this is the permanent structure of my game files. I'm still not sure everything is working after that. My latest build worked, and I did not dare try and dig any deeper. Never again. It took me OVER 9 HOURS.

• Overhaul of the game optimization to prepare it for the future (when I have hundreds of CGs, etc...). Previously the game peaked at over 160 nodes, and now it sits around 40. I also put structures in place to limit the loading of CGs based on the scene. Rather than loading every CG in the entire game, it will load only what is necessary for the current scene. This does still cause a little bit of lag when the scene is loading, I might be able to fix that later, but for now I'll just say deal with it, because this isn't very easy to fix.

• After completing a phone conversation, you can now go back to the home screen and complete other conversations. Currently there is only 1 functional conversation, so this is pointless, but it sets up the game to proceed from where it left off.

I did not add any new content or CGs in this update, and there isn't really anything interesting to screenshot, so I'll post a preview of an upcoming CG that is not in the game yet:

Latest Download:

You must be registered to see the links

 

[NoTraceOfLuck]

One download!! Nice!!

I saw someone actually downloaded one of the previous versions, that's my first download. I understand nobody is downloading because the game is pre-release and has almost no content, but I'm still happy about.

Here is my update for today:

My current priority: I consider the story up to this point to be the "Introduction." My goal is to polish all the game systems as much as possible and make the intro feel really good to play. As well as build systems that will help with the rest of the game too.

So for now, I am only adding very limited content. However, there is some slight amount of new content in this update.

Changes

• The choice selection screen was modified and looks a little bit better
• You can now click an image (in either Wallpaper or Messages) to view it in full screen
• I added another unlockable wallpaper (over 200 images generated to get it right) and updated the intro phone message story a little bit
• Added unlocks into the phone conversations (Depending on which route you take, you unlock a different one) and added a bit more polish to make the intro phone convo seem a little bit more realistic.

Here is my current list for what I need before I consider this initial stage complete:

1. Add 2nd phone conversation
2. Restructure some of the intro story to make the 2nd phone conversation make sense
3. Audio (I say this every time)
   1. Background music
   2. Sounds when tapping around the phone
   3. Sounds when unlocking something
4. Some animations for the character art in the intro. I tried to do this before and failed badly. But I am much more familiar with the engine now, and want to give it another try.

I am really happy with how the game is coming together. It is starting to feel very fluid and very solid. I have spent probably a total of 100+ hours creating this at this point.

Download link:

You must be registered to see the links

Gif 1: You can select an image in the phone conversation to fulllscreen view it (+ sneak peak at the new CG)

Gif 2: You can select an image in the Wallpapers app to fullscreen view it

 

[dreamerquest]

Hey there from a fellow developer.
I really like your method of sharing the development here (which I'll totally steal). :-D

I downloaded your game to see how it's developed.
Note: I'm not the kind that read, so it's a little bias.
- The text reveal is a little slow.
- Some consistency issue with the background images (western house vs japanese house+streets).
- Pretty solid, doesn't crash or bug out.

Given how hard it is to make consistent character with AI, I think you have done quite well.

 

[NoTraceOfLuck]

Hey there from a fellow developer.
I really like your method of sharing the development here (which I'll totally steal). :-D

I downloaded your game to see how it's developed.
Note: I'm not the kind that read, so it's a little bias.
- The text reveal is a little slow.
- Some consistency issue with the background images (western house vs japanese house+streets).
- Pretty solid, doesn't crash or bug out.

Given how hard it is to make consistent character with AI, I think you have done quite well.

Thank you! I also noticed the text is slow, I plan to make it faster by default and eventually let you change the speed / make it instant if you want, good feedback! I was going to do this later but I think I will prioritize it as my next improvement for now.

Yeah the consistency is always going to be a struggle. When I put those first scenes together I didn't put too much thought into it, for some reason the AI really loves to create Japanese things. AI art consistency will always be an issue for sure. I just spent about 3 days generating literally thousands of images to get some consistent art for the next scene I'm working on. Out of 2000+ images I only got 3 or 4 that I will use. Which is great, but there is no way I will ever finish this game if I spent this much time on the art. So I'm just going to accept that sometimes there will be some inconsistency

Anyway, thanks for playing

 

[NoTraceOfLuck]

TLDR: No new code release today. I spent the last week learning AI art. I posted some AI images in the next post if you want to see them. The rest of the post is just about my AI struggles. If you're here for the game, you can skip this one.

Latest Update:

Spoiler: This week I learned

You don't have permission to view the spoiler content. Log in or register now.

You might be asking, "You already have some AI art and it looks pretty good, so why are you doing this?" Well the answer is really simple, and it's the reason we're all here: NSFW images.

Up until now, I have been generating art through 2 methods:

1. Midjourney for fullscreen background art
2. AnimagineXL 3 for the character art

And Pika for videos, but on a limited scale. Pika is kinda shit, but there aren't many video options that are easy to use.

Source of the problems:

1. Midjourney is awesome, but it does not allow NSFW images at all. Period. The most you can get is a bikini pic, and even that requires a lot of effort to get past their censor. It's also too expensive to use Midjourney to generate huge quantities of images.

2. Animagine makes really amazing art, and it doesn't have a censor that prevents you from generating images, but it is still extremely hard to get it to output a NSFW image. Eg: You type in "Nipple" or "Nip slip", you literally have to generate 50 images to get a single one that has a nipple in it. And for a prompt like "Removing bra" which results in a much more complex image than just adding a nipple, just forget it. Animagine won't do it.

Considering my game is about poker, I am obviously going to have a strip poker scene (or multiple), so being able to generate images with a prompt like "Removing bra" is essential. Animagine won't work.

Another problem:

So why not just use a different model to generate the NSFW stuff? I tried that with many different models. Here's the problem:

All models output pretty different art styles. You can tune them a little bit, but you're never going to get eg: Midjourney to output an image that looks exactly like Animagine. So if I generate all the characters in Animagine, it is going to look really shitty when it suddenly shifts to a completely different artstyle for an image from Pony. Also, nothing looks as good as Animagine or Midjourney.

Choice I Had To Make:

I have a few options:

• Just say "fuck it" to having a consistent art style. Produce NSFW images in Animagine when I can force it to, and fall back to a different model otherwise.
  • I tried this, but other models just suck compared to Animagine. They're so bad that I don't event want to work on the game anymore if the art is going to look that bad.

• Find a model that looks pretty good and allows NSFW art more easily, and just convert the whole game to that (for consistency) and accept that the game is going to look worse.
  • Doing this would make me just give up on this game. I really really want to continue using Animagine for my SFW CGs because they look bomb AF.

• Figure out some way to make the NSFW CGs look as good as Animagine (or at least, stylistically consistent with Animagine). This requires delving deep into the depths of AI art and learning a million different complex things.

Option 3 Chosen

I decided to take a crack at really learning AI art, rather than just sending text to a model that has the default configuration. I did not expect this to be so hard, and I actually quit 2 times in the first 48 hours, almost completely giving up. But then I found

You must be registered to see the links

which finally made everything start to come together.


Get ready for my next post, where I will post about my journey, with images, and the final product.

 

[NoTraceOfLuck]

The AI Journey (lots of art)

I doubt anybody will ever read this whole thing but some people might scroll past it and enjoy the images so I'm going to post it.

This is journey to solve the problem I mentioned in my previous post: I want to use Animagine for CGs in my game, because it looks amazing, but it gives a really big challenge generating NSFW art. For this reason, I dove deep into the world of AI.

My Goal
I needed a process that will allow me to generates images that 1: Look artistically consistent with Animagine, and 2: It can't give me a huge fucking hassle generating NSFW art.

My prompt:

• Generate my character, in a coffee shop, leaning forward, with a downblouse.

I chose this prompt because a downblouse a relatively complex image. Not too simple. But also not too crazy. If a model can't do a downblouse, it won't be useful for anything else in the future.


Reference Art(Animagine 3 XL)

All the rest of the art in my game will be generated using Animagine. So I need my final output to look similar to Animagine art. This is the Animagine image I will use as a reference:




First Attempt

I spent a ton of time trying to understand Stable Diffusion. There are 2 versions people use. SDXL and SD 1.5. SDXL is newer and better (Animagine is based on SDXL) but 1.5 has a much bigger community with more plugins. So I decided to start with 1.5. I used a model called Counterfeit 3 which is based on 1.5.

I put in my prompt, and here's what I got:



Wow. Holy shit. I almost quit immediately. It looks fucking horrid.

I did a ton of work adjusting the prompt, adding plugins, etc... And I got it to output something I actually liked:

And I wanted to test out more complex prompts, such as having her be embarassed and cover her chest:



It worked pretty good. I was happy with it. But there's a problem here. It looks absolutely nothing like the reference image. Also, the backgrounds are really ugly compared to the reference image. The background is also ugly and yellow and every image I generated seemed to have this same dull background. I almost stopped here and just decided that this would be good enough.

Enter Pony

But then I started to read about Pony. Pony is an SDXL model (not 1.5). SDXL can produce higher resolution images, and this can fix the ugly backgrounds and overall result in higher quality art with more detail.

That sounds pretty good. So I loaded in Pony and generated an image:


Holy fuck this looks awful. Absolute barf. You can see the backgrounds are much higher quality, but the actual character art is so so terrible compared to the reference image. There is no way I could use this.

I did a lot of work on the prompt and locating plugins, and I got something that looked pretty ok:


But it still just really doesn't look like the reference image. Also, this image was by far the best out of many really really bad ones. I wasn't feeling too confident on Pony.

But I had an idea. I was using the name of a particular character to generate these images. However, you can see in the reference art, the actual character isn't that noticeable. If you know who it is, you can kind of tell, but the hair and stuff really are pretty different.

So my idea was to stop referencing this character directly, and just try to reference "An anime girl with short brown hair." This is what I got:



Honestly this looks pretty good. But it is not consistent with the reference image. Completely different art style. I could probably accept this, but I wouldn't be happy about it.

Pony Checkpoints

That's when I learned about something called Checkpoints. These are models that were trained on top of another model. Usually somebody does this to keep most of the original model, but produce a stronger style (eg: more Anime-ish rather than more Lifelike)

So I downloaded a few Pony checkpoints and got to work.

The first one I downloaded was called Autism Mix:



The art style for this one is really beautiful, and it looks really good, but it is not consistent with the reference image. Also, the other images generated for this one had issues. Overall I might have used this one if I didn't find anything better.

PD For Anime

Then I found one that sounded really promising. It was called Pony Diffusion for Anime (PD For Anime) and it is supposed to be a checkpoint of Pony but with a strong anime art style.

I gave it a try:



This looks both really good and really bad. It looks good because it is the closest thing to the reference image yet. It has an anime style face with flat features. It also didn't complain about the downblouse. But it also looks terrible because the colors are just awful and oversaturated and burnt.

I did what I could to reduce the saturation, and ended up with this:



It looks better but really still not acceptable. Also the backgrounds seem to have lost all detail.


Confetti

I found one called Confetti. I started it with the "Brown hair girl" prompt rather than referencing my specific character. This is what it gave me:




This looks actually pretty good. Stylistically it's great and it's very consistent with my reference image. Also, the backgrounds have pretty good detail. I like this one, but I did not like how I've completely lost the original reference character. There's no way you can tell which character I'm referencing from this image.

So I tried adding the character's name back in to the prompt:



And it just looks really really bad. Just not what I'm looking for. At this point, I'm considering sticking with "brown haired girl" and just losing the character reference. But I have a few more to try


7th Anime PD




This one actually looks awesome. However the art style still doesn't match very well with the reference image. And this model also generated much more of a "painted" style overall, resulting in some of the other images that it generated just not looking great.

I also noticed there is no nipple in the downblouse here. This usually means the model is fighting me. I can't remember exactly if that's why I scrapped this one. I do still like this style and would use it if I could get nsfw generation to work without pain.


Mugen

At this point it's like 3am and I'm tired and this is the 10th model I've done. Each model requires a lot of work and adjustment to try and get it at its best. So I load up this new one called MugenMix or something like that.



Wait... wow. That looks extremely similar to the reference image. It kept the character reference almost exactly to the reference image. It looks a little bit burned and oversatured but I can deal with that.

I need to figure out if this is the one. So I tested the NSFW filter with something more explicit:



Wow. I really like this. It gave me zero problems generating NSFW art. But it still has some issues. First off, the face doesn't match the reference image very well. And overall it still looks oversaturated. I did a lot more refinement on this model and came up with this:



This looks really good and it's almost certainly going to be the model that I use. Not perfect, still kind of shiny and oversaturated, but much better than every other model.


Something New - Animagine + Pony Mergers

Then, in the early depths of the morning, something catches my eye. I see a model called Reponygine 3. I read that it is a merger of the Animagine 3 model with Pony XL. Could this solve all my issues? Animagine has the art style I love, but Pony is much less strict about NSFW artwork.

This could be it, I gave it a try:



And I generated another:



This second image was very important. Some models are really bad at understanding complex instructions. "Removing bra" gets completely discarded by Animagine, and other models also struggle with it. This model understood it perfectly and generated this image on this first try.

This looks visually really good and really consistent with Animagine. But it had 2 major issues: the first is that it kept generating lolis. My reference character is a full grown adult and so this would not work. The other issue is that it really just lacks detail and the lighting is unimpressive.

So That Was It

Mugen generated some great art, better than everything else, and that's what I decided to go with... wait. What's that?

Refiners

This is when I discovered what is called a Refiner. A refiner is a special step in the generation process where 1 model generates the first part of your image, and a 2nd model completes the image.

This is exciting. Can I use Mugen to generate the NSFW concepts of the image, and then force Animagine to actually fill in the art? This might be a way that I can literally force Animagine to draw NSFW art, because it needs to complete what Mugen started.

This requires weighting the models. Eg: Tell it what % of the image Mugen should generate first, and then Animagine will finish off the rest.

I give it a try:



This looks good. But not great. The face does not match the reference image, and it looks a little too 3D. I decide to tone down the % of generation for Mugen and given Animagine more time:




WTF IT ADDED A SHIRT. This is what I'm talking about. Animagine just does not want to generate NSFW art. It literally added a shirt into the image even though my prompt was just things like "Topless, no shirt".

This was annoying, but if yous top for a second and look at the photo, you notice 2 things. 1: The background art looks great. It has the quality of Animagine's backgrounds.

And most importantly 2: This looks extremely close to the reference image. Scroll ALL the way back up to the top of this post and you will see it is almost an exact match. That's exciting.

So I start tweaking things. I need to give Animagine as little time as possible to put a goddamn shirt into the image. I need to get Mugen to generate almost the entire thing, but leave just enough for Animagine to fill in the face details, but no shirt.

And I finally get it. The perfect mix:



Perfect facial match to the reference image, perfect art style match, beautiful background, and fully NSFW. This is it. I finally got it.

It's not perfect. Actually, this image generated with no nipples. Just blank tits. I had to use another step in the AI process to generate the nipples (called Inpainting). And that's a little bit concerning. Because it means I will still be fighting with the AI in the future. However, I can always fall back to Mugen alone if needed.


Thanks for reading (or just scrolling through the art), I hope you enjoyed it.

Updates to the game will proceed in the next few days. None of the art depicted in this post will be a part of the game.

 

[NoTraceOfLuck]

I promised audio like 5 times and it's finally here! (Sort of)

Latest Changes

• Added background music
• Added sounds for all functions in the phone (messages received, sent, etc...)
• Increased the text display speed. It is not adjustable yet, I plan to add adjustability when I get around to adding a proper menu system
• Added some basic animations to certain character art transitions

What I'm Working On Next

• The audio is "fine" but it's really just 1 song on loop. I'm adding more than 1 song and will make it change based on the scene you're in.
• Weekend project will be one of these 2 larger efforts:
  • A proper menu system
    • Saving
    • Loading
    • Adjusting Volume
    • Adjusting Text Playback Speed
  • OR adding in the necessary framework needed to progress from the Phone scene into the next main component of the story.

Adding a menu system itself is easy, but handling saves and loading will not be easy. This will likely not be a "Load / Save the game anywhere, any time" system. It will instead allow you to load checkpoints in the story. That will be much easier for me to do. Still difficult, but easier.


Previously I said I was going to attempt to add the 2nd phone conversation with the Poker Girls, but I think I need to do some larger planning around how that will work. Originally my plan was to have only ONE character so I didn't really plan for adding them to the story, I don't have any character art for them, and I generated them in Midjourney so it will be hard to generate similar art styes if I want NSFW art of them later. So overall, I'm not sure yet on how I will handle them.


There isn't much to screenshot this release since I can't screenshot audio, so I took a small gif of the character transition animations I mentioned

Download:

You must be registered to see the links

 

[NoTraceOfLuck]

Big Weekend Update

Latest Changes:

• More and better music
• More and better sound effects, especially for the phone
• Emojis in texts appear slightly larger font size than the rest of the text, to make them easier to see

FULL MENU SYSTEM

• Ability to adjust text playback speed
• Ability to adjust background music volume
• Note the menu is ugly as hell, but I don't care. There are certain parts of the game where I put a lot of craft into the beauty of the game, but the menu system will not be one of them. I found the whole development of this system to be quite boring and I don't want to work on it any longer.

SAVE SYSTEM

• Ability to Save
• Ability to Load

These are a lot of changes that are not very exciting, but adding saves was a ton of work. Likely there are still scenarios that I have not properly covered. This is one of the biggest drawbacks of using Godot instead of Ren'Py, having to create my own save and loading system. But it works pretty dang good.

Known Bugs:

• One new bug was introduced in this update, which is that you can select 'Load' on an empty save. This does unexpected things and you probably shouldn't do it. I may fix this, or I may not. I haven't decided yet.

What's Up Next:

• I am pretty quickly approaching what I would consider a release-ready candidate. Version 0.0.1. I need to finish up the following items before I am release ready:
  • Ability to progress from the Phone into a stage following the phone. Currently I don't have the necessary foundation in place to move from the phone to another scene, so I just need to build that out
  • Add some additional flavor to some of the images by using AI to make small changes to the image (change facial expressions, maybe slight arm position movements, etc... very small things that just add a little bit more life to the scene.)
  • Add another possible modifier to Character Art. Currently I have a sort of "spin" effect at some points in the conversation, but I would like to add a 'Zoom' effect in some cases, I feel this can add some good flavor to the game.
  • Possibly: Put a little bit more life into the phone UI. (Eg: Add character face art to the messages screen, etc...) Just see what i can do to make it look a little bit more full of life.
  • A 2nd text chain for the Poker Girls, possibly with a new CG.

As you can see, there's not much left that would be considered a major effort at this point. Just some flavor enhancements and a little bit of new content. Version 0.0.1 will likely release with only 10-15 min of content at most, but with this foundation, I've created a system which allows me to add new content quickly.

Latest download link:

You must be registered to see the links

Screenies:

 

[NoTraceOfLuck]

Hey all, I'm hard at work on the new update, but I wanted to share some new AI functionality I am working into the game. This stuff is incredible. Take a look:

View attachment watermarked_video007da6678c8de4de5ae474bdc42a81d11.mp4
View attachment da7852fe-4fc5-4606-9501-2fabd5106a72_raw_noprompt1_video0f0101754bae1451e8fbeb8b2c2a10482.mp4

 

[NoTraceOfLuck]

New character who dis?

Latest Changes:

• New Character!
• More flavor for the phone
• Rework of intro dialogue sequence to introduce new character
• New cgs and character art for new character

What's Up Next:

• A 2nd text chain for the new character, possibly with a new CG.
• Ability to progress from the phone into the next sequence
• I probably need to start compressing some stuff because the game really should not be 500mb for the amount of content it contains

Latest download link:

You must be registered to see the links

Screenies:



Phone updates (more interesting texture on the background + added character icons)

 

[NoTraceOfLuck]

This post is not an update (see the previous post for the update I released only a few hours ago), I just wanted to provide some more info on my AI process when I have been working on for the last few weeks to get better and produce better art.

I primarily use 3 tools for generating art. The first one is Midjourney.

This tool is a simple one, put in a prompt, and it outputs an image. You have the option for "varying" the image if it is close to what you want but not 100% correct. Sometimes I vary an image many many times before I get what I want.

Midjourney produces excellent, really beautiful images. And it has very strong 'prompt adherence' which means it will actually understand complex prompts with many actions in them.

The problem with Midjourney: It just does not give enough control. If you want an exact facial expression or hand positioning or very fine-tuned control over an image, it is really hard to get it with Midjourney.

And it is 'PG-13' only. No nudity allowed.

What Midjourney is great for: 2 things

• Generating images that don't require exact structure. Eg: landscapes, settings, etc... I use it for pretty much all images that don't focus on a particular person as the character

• And 2, it's great for generating a base that is very close to what you want, but not perfect

And so my 2nd tool I use to solve the problems are Midjourney is called InvokeAI.

Invoke is a full studio for generating images using Stable Diffusion. It gives almost unlimited control over the outputs.

The problem with Invoke:

In Midjourney, you literally type in some text and click 'generate' and you're done. It's easy. In Invoke, there is a lot more work to be done for each image. You get a lot of control, but it comes at the cost of higher effort.

And the 2nd problem: Stable Diffusion can produce amazing artwork, but it can also produce really ugly results. While Midjourney is beautiful every time, Stable Diffusion has a wide range and can easily result in ugly, burned images. There is a lot of time spent looking for models, looking for other tools to enable all the functionality and art styles.

When Invoke is great:

• When I need absolute control over an image

• When I need nudity

• When Midjourney generated an image with great structure, but the art style does not match the rest of the game (eg: see the above screenshot of Midjourney, there is a series of multiple different art styles. Some will not match my game) I can use Invoke to change the art style.

See an example here where Midjourney output this image:



It looks great, the structure is perfect and the facial expression is top notch. But it just doesn't match the rest of the game. And also, it just doesn't look that good.

So I can use Invoke to manually change the art style of this image, which I did. I changed it to:


Which I think looks a lot better and more consistent with the rest of the game and especially with the environment that this scenes takes places in within the game.


3rd Tool: HuggingFace Inference

This last tool is kind of a special one, I won't write too much about it. There is a website called Huggingface which hosts a lot of different AI models. Huggingface has the ability to "inference" against AI models. Which means they host and run all the actual generation on their servers, you just give them a prompt.

If you read my previous post on AI art, you know I am using the Animagine 3 model to produce my character art. Hugginface hosts this model for inference.

And so I have a special script for this website which basically takes 1 prompt (and some variations of the prompt: eg, I might vary the emotions etc...) and basically runs a Huggingface inference as many times as I want. I often set this to 500+ times to generate images.

So I basically use this tool to generate huge amounts of images in the hope that I can get a few that I actually enjoy. This gives me ideas, inspiration, and sometimes actual usable artwork.

See this example where I basically wanted to generate a "casual night" with one of the characters. I used Hugginface Inference to output hundreds of this prompt so that I can get some inspiration on how I want it to look:

 

[achmednasir246]

i really like your art. even though it is ai generated one can see the amount of work you put into it to make the images express your creative idea. i think while you are not drawing the images yourself it is still a huge achievement what you are doing.

I notice all of your images are in portrait mode. Is this because of a choice within your game (i.e. have it played only on mobile) or is this a limitation of your ai process?

 

[NoTraceOfLuck]

i really like your art. even though it is ai generated one can see the amount of work you put into it to make the images express your creative idea. i think while you are not drawing the images yourself it is still a huge achievement what you are doing.

I notice all of your images are in portrait mode. Is this because of a choice within your game (i.e. have it played only on mobile) or is this a limitation of your ai process?

Hey thank you for the comment & the kind words! I actually enjoyed thinking about this question a little bit, so I wrote more than you probably wanted:

TLDR: It is a limitation of my AI process, but not just because I can't do other resolutions, but mostly because I just like this one the best.

When I first started, I used an online image generator for my art. And that generator had 3 options: 1024x1024 (square), 896x1152 (portrait), or 1152x896 (landscape). I thought those were my only options, and I thought portrait looked the best considering the small sizes. That was why I originally picked it. However with my new tools, I can generate images of any size. But I still choose to keep this resolution for a few reasons:

1.
AI is really good at generating either a character/foreground, or a background. But not both. In my title post, look at the first 2 images. The first is a really gorgeous outdoor scene with orange trees and a ton of detail. But no characters present. The second image has a character, but notice how low quality the background is. It's very hard to get both a beautiful background and beautiful character on one image.

In portrait mode, it isn't as noticeable, because the character art can fill the whole image so there isn't as much ugly background. I generated this image right now in 1920x1080 (it's not great, I'd never use this, but it's a good example). Notice how in this resolution, it gives so much more space to the background, compared to one in portrait mode:




and the 2nd reason:

2.
The #1 biggest issue w/ using AI art for a game is consistency. When 90% of the screen is taken up by the character (in portrait mode), the only thing you really need to control is the character. But if 50% of the screen is your character and 50% if the background, you have to control both the character and the background, which is very difficult to do.


And 3. I just like that it's a unique resolution. I think it's a little bit more memorable and a little bit more unique for people to play with.

 

[achmednasir246]

hey thanks for the answer. especially the second reason is something i also encountered.
did you ever try to use inpainting (like with your nipple problem) to combine your detailed characters with a detailed background?

for example:

create a great background with some tool (sorry for the artefacts, had to rip out the girl from your example ):


create your detailed character:


and then inpaint it to the backgrounds (sorry for not fixing the hands for this example):


or with a different character image:

 

[NoTraceOfLuck]

hey thanks for the answer. especially the second reason is something i also encountered.
did you ever try to use inpainting (like with your nipple problem) to combine your detailed characters with a detailed background?

for example:

create a great background with some tool (sorry for the artefacts, had to rip out the girl from your example ):

That's a good idea, I think it's definitely possible. I did a test here with a feature called 'regional guidance' which basically allows:

1. A main prompt which generates your background
2. A 2nd prompt which only applies to a small part of your image (which you can use to place the character)

And it came out with a really beautiful background, but the character still has a lot of issues (really long legs, and white hair for some reason):




Here I will try with normal inpainting, I start with a beautiful background:



I draw a shape to represent how the character should look:



And it does work pretty good (a little bit weird, because it is trying to match my drawing, which wasnt very good):




It definitely works. It would require a little bit more work to make the character look less deformed. But I also want to be sure I don't put too much time into 1 single image, since I am just 1 person I need to split my time between the game development vs the art.

Good suggestion, I think it looks good.