Unity Heart of the Tower [Dev Thread]

[Sentinel Sparrow]

Testing a movement system
i can move 10k enemies without dots on the main thread ~50fps PC (via grid cellular automata)



i want to be able to render far as well so this could work for a voxel tile based visuals, it feels like a stepping stone.
The goal is ~3k enemies and to run efficient for mobile.
Maybe for arena's there will be more focus on the enemies and walls or obstructions for distance culling allowing more enemy units.
I do want to increase render distance to see far off points of interest when platforming. A grid based voxel system might allow this and ease of level generation.

 

[Sentinel Sparrow]

my game will support Ultrawide monitors QHD (3440x1440).
i did a test i have 10k enemies working at a solid 100fps, im not using burst or multi threaded code but based on my cpu usage its doing it lol
View attachment 27 - 20k grid movement UWQHD 120fps 2.webm
Also im getting 30fps on mobile with ~7k enemies (1440 x 2960 on a Qualcomm Snapdragon 845)

- updated frame times and enemy units, this amount is a best case scenario and will probably decrease depending on level assets, projectiles etc
- reworking the character controller to feel weighted, need to add the new input system, fix mobile movement

 

[Sentinel Sparrow]

hey guys i updated my first post, what do you think of the format? any suggestions?
I plan to add a proper image banner once i have the goddesses art style figured out. twirls moustache

 

[DuckHeal]

hey guys i updated my first post, what do you think of the format? any suggestions?
I plan to add a proper image banner once i have the goddesses art style figured out. twirls moustache

Looking good so far. Seems like you've decided to go big for the first release

 

[Sentinel Sparrow]

Looking good so far. Seems like you've decided to go big for the first release

Yeaahhh ^ ^; i like building systems, definitely taking my time on them. i want the foundation solid so the game can grow naturally.

Content size will depend on time and support, of course, but the game is fully free during development. All builds will be uploaded through Patreon’s free tier, so anyone can follow progress without paying. but this is the expectation anyways.. just grasping this concept now.. /=.="

 

[Sentinel Sparrow]

Hey what do you guys think of my writing thus far?
Its how the game starts.
Eventually when i have the time or money to hire an animator ill animate it. ( * w * )

Spoiler: HOTT Story: Prologue first half - WIP Excerpt

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

 

[Sentinel Sparrow]

been working on the story, and realized i might need to make AI behaviour for a companion to make my story work.. i want a romantic love story between the player and them. which for a 3rd person combat game is hard to get right. most likely will end up looking like a confused roomba... =.="

which makes me think if im going to have an Ai companion that maybe i should just make low poly characters of them to represent in game. i can reuse the model for action poses instead of drawing every action pose like with 2D VN UI. this would allow the characters to be grounded in world space and simplify the cinematics (instead of 2D manual animation; my laptop cant handle ai rn and the consistency is poop). Im gonna need to 3D model some of the environment anyways... /sweating/


AhhAHhh my brain rn

But then this creates a discrepancy between my current 2D enemy style and 3D characters, so maybe pre-render the 3D characters?... or just make enemies low poly 3D as well.... /face grating indecision continues/

If i go 2D drawn I would have to mange the facing direction of the player and the companion for every interaction action, hard to visualize but i think possible. depends on how many directional movements the player has. keeping it to 4 the cardinals N S E W can it make it easier.

this isnt the style, its just a visual ref of the cardinal directions i was talking abt.

I also need to link the gameplay mechanics to the story themes:
- Reason for resurrecting.
- Reason for AOE Vampire Survivors like attacks and spells.
- Possible reason for companion only showing in cutscenes.

The ideal would be
- VN portraits for poses in world space rather then VN UI, and it just works, pose actions match up with all possible directions.
- facial portrait expressions ✅ def use this
- have the companion show only for cutscenes...me thinks ... ✅ most likely if 2D
- have them static poses then animate them... (ideally animated)
- still having 2D sex scenes like those from Spline 2D etc ✅ certain

 

[Sentinel Sparrow]

dam.
this is what i spent a year trying to figure out/make. they even sketch then use ai to finalize the character drawings keeping creative IP rights saving 90% of the time spent on the last 10% polish. The animations are sloppy though barely effective. They refined their art pipeline to maximum efficiency with minimal effort, even using low poly 3D for characters and 2D minimal tiles for environment floors, walls and props like seaweed. if anyone is serious about making money from curating/creating their own adult games then mimic them. ai handles the writing and reads all the same, the polish from drawn characters can be tweaked to look less like ai plastic. the low poly models work for depth and rotation (especially for cinematic cutscenes if when they decide to use some of their 18k CAD / month to spruce up the game). the fixed camera angle works to keep 2D aspect of it less noticeable (2D VN static poses and level props). Hats off to you Tiger Roll Games, i wish i had this foresight a year ago. hindsight is 20/20 for me.

while kicking myself i see a lot of areas of improvement but most improvements cost more time. theres basically no more corners you could cut, bravo. once this becomes the norm then you'll have to use a small portion of your fortune to nudge yourself above the rest like nintendo/pokemon. (ofc this is after tax man blows up your insides, your dept collectors bounty paid.. and your new cars, ebikes, roomba, sex bot, bar table, stripper pole, strippers so you can get ntr by your sex bot.... sheesh no wonder the game is the last thing paid for )


non nerds are usually the ones who pay, convenience sells and it shows they have 2.5k patrons. is it the concept, is it the genre, or is it the gatekeeping... seems to be all 3.

gentlemen and the one off lady who stumbled in, get ready for this wave of rogue like hybrid porn games. /recedes to my overly engineered game that will take 100x longer to only turn out marginally different with the same feel of ai plastic taint that fewer to none care of/

so TAHOP is p cool, the game shed some light as to what i can do for art workflow.

I decided for my game characters will be 2D illustrated; which is dumb and will take more time then low poly 3D, but i already made my 2D sprite render shader optimized SRP shader. it can also look cool like Cloud Meadow if done right. i picture it looking like a modern 90's anime style (*w*)

For the 2 main characters i'll cut corners by limiting the directions. All other characters will be either quarter pose or new poses required for sex scenes and possibly cinematics (though i doubt ill have the time to do animation for cinematics).

1. quarter pose front facing (this will be mirrored/flipped for the left facing)
2. front facing (towards the camera)
3. back facing (this can be a black silhouette of the front and flipped, most likely ill need to be drawn)
4. quarter pose facing away

so thats 4 new drawings for each animation, each animation can be like 4-20 frames.
im thinking about adding an affection system where they can do nerdy and naughty things

 

[Sentinel Sparrow]

Gameplay decision update

i forgot about the MC armor sets, weapons. was planning on having light heavy attacks.

but now im thinking i should only stick to Vampire Survivors like combat, aoe + projectiles rather then melee attacks...

which sucks for meaningful combat but saves so many drawings
(weapon stance x armor x base character 4 directions x animation frame, for each new armor and weapon type, each animation...).

sigh i should really switch to 3D model if im going to have weapon and armor change, but my 2D pipeline i built and its optimizations would be for nothing ;-;

i could make each direction pose like a 2D skele rig instead of a drawing but the animation would look blocky.
maybe use a 3D base character model for directional perspective, sketching over then ai to polish it.


im thinking maybe just have armor and items shown as orbs on the ground, and only visible in chest and inventory menus. so the D2 loot goblin in me still gets sweet looking loot
]

maybe player visuals only change by skin recolor, auras and possibly tattoo's that glow. if at all.


for items i do like roguelike games that uses gun attachment mod system, i could only have weapons and no armor so the player avatar doesnt need to visually change allowing for light and heavy attacks but i really like it when the player avatar updates from armor equipped T-T


so the perma growth is your equipment, and in game you start from level 1 each floor to x level like VS. this keeps the extraction risk reward. i also want to add an item orb upgrade system like PoE for sweet fine tuning, loot goblins rejoice.

 

[Sentinel Sparrow]

Art Style mood board
art direction / styles i'd like to capture in HOTT

Characters (90's modern anime) dare i say it, yes, yes you can fuck these



Enemies (Dark Fantasy / Fantasy) no, no fkin these




Environment (if i can get streamable world working i'd like to render far like these)



alternatively i could make the enemies all monster maidens / monster girls. this i debate to not do because sex scenes will take a lot of work. so its why i gate it to only so few main characters. I'd rather have good sex scenes then sex scenes everywhere which is the crux of the problem all porn game devs face forcing them to make a hybrid porn game (at least for me).

the games turning out to be ram hungry so we'll see if the art turns out more pixellated.

hmm.. if the world monster be like this
and the main characters are like this ... then i should probably update the monsters to fit the style as well sighhh and the quality time goes up... theres no winning. fly you fools. like playing an anime.wicked. but ill never finish. well oobviouuusly learn 3D.fk im so done i just wanna make this thing.

 

[Sentinel Sparrow]

pushin forward
i tried comfy ui and its p simple n doesnt make my laptop go turbo granny.
fingers crossed it doesnt explode my laptop charger like last time, 6 months ago...


art workflow V22222.2 to get somewhat consistent characters:

1. Hand draw the base pose (paper/photoshop from refs)
2. Clean the lines with Comfy UI
3. Vectorize it in illustrator + fill in flat colors
4. Polish by adding shadow + highlight with Comfy UI
   * try frame pack = video turn table for new angle references???

* need to figure out how to use lora's and input images in comfy ui to make this work. specifically the line art control net?

Time to define the character visuals before i can run this workflow test:

1. decide the looks for 3 main characters
2. gathering reference images
3. sketch the main quarter pose idle stance
4. sketch expressions
5. build the comfy ui nodes
6. test the comfy ui workflow

dont forget, make things bc YOU want to see the ideas come to life;
money follows passion, passion follows authenticity.
a small reminder

and dont stress the small things. lock it in and keep going.

 

[Sentinel Sparrow]

holy fucking cockles

If you have a google account you can use gemini and in gemini you can use nano banana. Results generated from nano banana is restricted by licensing so most likely you can only use them for references. I think 4 images per day for nano banana pro and unlimited for nano banana.

If you download the latest version of comfyui you can use qwen easily. Comfyui has everything set up for you. No cost.

Qwen can do nsfw.

The "qwen" I mentioned on the above is "qwen image edit". Not qwen the LLM.

dude thank you for mentioning qwen


----------

edit: so its not that good, the image to change needs to be well lit and generally normal proportions (specifically for women). The style also changes based on qwens model, and new things are added and removed. i guess i got lucky my first try.

i see this useful for getting the outfit and character into a dynamic pose, which you can then trace and fix the details manually... or fix it up in photoshop. in the case of my bunny maid frame pack worked better as a turn table, but using qwen character pose in comfy ui allows for dynamic poses that need A LOT of fixing /sigh/


------------

Spoiler: lmao

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

 

[Sentinel Sparrow]

I had an idea to make the levels out of primitives 3D shapes, like kit bashing so i can work quickly in unity (from swords to walls and arches). but then realized its inefficient (on a scale of 10k things) because unity processes a lot for each game object:

• Transform hierarchy evaluation
• Local → world matrix calculation
• Bounds update
• Culling checks
• Renderer submission
• Component iteration

so reducing objects, combining meshes is the optimal way to go (for large scale scenes).

ill still use primetives to greybox the levels.

3D animation optimization on scale = calculating the stuff before so it doesnt at run time. this means saving each mesh in its animated key frame pose (vertex positions) and swapping them like a 2D sprite sheet / flip book. this saves on skinned mesh renders which cost in game object transforms and draw calls.

here is a method for saving simulation meshes without bones into textures O.O

i noticed when learning 3D its best to search for exact tutorials on the things youre building instead of dull boring intro tutorials.

incredible swords tutorial (so many advance techniques that can be applied in all modelling):

female characters (though i have a feeling theres better tutorials out there, kinrean know of any good ones? dude check out the sword one)

 

[kinrean]

for female model, i can't say i am very good at 3d model, but the key in poly number or location is all for the animation or how it move.

what i mean is that to know if is good retopology/ mesh is have to see how you move it for animation. like the breast, may seem simply but every one make it different by a bit because of what we need for breast movement or breast scaling is not the same.

and the higher the poly the more easy it is for animation. that is why is harder to make low poly model.

so in short is there a tutorials i will recommend you, not really as all about the same , but if you want to look like anime you need to learn the face normal .

You must be registered to see the links

, and i will recommend you to know or see how it move.
and that is why the video you show recommend you to download other people model to learn ,and i will say not just how it make but also how it plan to move. and is a skill to learn from other people too.

and i do warn you the hand/ finger , every one make very different for many many reason. that i really think you should copy as much as you can.
i have seen a old Resident Evil model, the hand , they make 2 set of hand, 1 set for normal walking (open hand), another set for holing gun. by doing it this way they can make the hand look nice and no bone for hand (2 hand bone is more bone then a whole body), save cost and look nice.

and in truth, i think you should just take a base mesh model and retopology it. as alot people do it this way to save time, faster then using 2d image.

 

[mibc9394]

holy fucking cockles
View attachment 5520100


dude thank you for mentioning qwen


----------

edit: so its not that good, the image to change needs to be well lit and generally normal proportions (specifically for women). The style also changes based on qwens model, and new things are added and removed. i guess i got lucky my first try.

i see this useful for getting the outfit and character into a dynamic pose, which you can then trace and fix the details manually... or fix it up in photoshop. in the case of my bunny maid frame pack worked better as a turn table, but using qwen character pose in comfy ui allows for dynamic poses that need A LOT of fixing /sigh/
View attachment 5520412 View attachment 5520415 View attachment 5520420

------------

Spoiler: lmao

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

Where did you get the qwen character posing workflow? I wanna give it a try

 

[mibc9394]

I had an idea to make the levels out of primitives 3D shapes, like kit bashing so i can work quickly in unity (from swords to walls and arches). but then realized its inefficient (on a scale of 10k things) because unity processes a lot for each game object:

• Transform hierarchy evaluation
• Local → world matrix calculation
• Bounds update
• Culling checks
• Renderer submission
• Component iteration

so reducing objects, combining meshes is the optimal way to go (for large scale scenes).

ill still use primetives to greybox the levels.

3D animation optimization on scale = calculating the stuff before so it doesnt at run time. this means saving each mesh in its animated key frame pose (vertex positions) and swapping them like a 2D sprite sheet / flip book. this saves on skinned mesh renders which cost in game object transforms and draw calls.

here is a method for saving simulation meshes without bones into textures O.O

i noticed when learning 3D its best to search for exact tutorials on the things youre building instead of dull boring intro tutorials.

incredible swords tutorial (so many advance techniques that can be applied in all modelling):

female characters (though i have a feeling theres better tutorials out there, kinrean know of any good ones? dude check out the sword one)

If you want to have 10k moving characters in your scene (I assume that's what you want to do?), what you need is probably VAT (which you already mentioned) + GPU culling + (maybe) a custom LOD switching mechanism.

VAT is vertex animation texture. It allows you to move a character with a texture, so that the calculation is completely on the GPU, which is good GPU instancing, so that you can have many of the same character in a scene. And with some shader setup, you can have variations.

But for that amount and scale, solely using GPU instancing is not enough, you need to implement culling logic so that what's not visible in the camera won't render. Since GPU does not do culling by itself you need to implement the culling logic yourself.

There should be some plugins for VAT and GPU instancing with GPU culling available.

If the above are not enough for the performance, then you probably also need to split up your scene into chunks using some sort of spatial partition algorithm and only calculate the chunks that are close to the camera.

And then for making large scene it's always about how you pack all the data together and reuse the same things as much as possible (same materials, same textures, same models, same animations etc.)

 

[Sentinel Sparrow]

kinrean thanks man, solid advice. especially for the normal tutorial and hand swap and ref model is excellent planning the right topology and normals for the desired animation is something ill keep in mind do you have any good base models in mind for anime characters? i remember Guilty Gear XRD had some amazing looking anime style normals


mibc9394
here you go bud may it be more useful to you then for me ^ ^

You must be registered to see the links

i used the comfyui manager to install missing files

I do have a culling system based on AABB camera frustum planes check, its not perfect the images on the side get popped in but it allows me to get good fps right now. the VAT method looks promising as well! Yeah streamable assets in unity is on my to do list for the level assets ^ ^ thanks so much for info

 

[mibc9394]

kinrean thanks man, solid advice. especially for the normal tutorial and hand swap and ref model is excellent planning the right topology and normals for the desired animation is something ill keep in mind do you have any good base models in mind for anime characters? i remember Guilty Gear XRD had some amazing looking anime style normals
View attachment 5537653

mibc9394
here you go bud may it be more useful to you then for me ^ ^

You must be registered to see the links

i used the comfyui manager to install missing files

I do have a culling system based on AABB camera frustum planes check, its not perfect the images on the side get popped in but it allows me to get good fps right now. the VAT method looks promising as well! Yeah streamable assets in unity is on my to do list for the level assets ^ ^ thanks so much for info
View attachment 5537660

For the popping, just reference another camera that is parented under the main camera. Then set the frustum to be a bit larger for the referenced camera and disable it. This is an easy way I found to solve the popping issue. No performance increase at all as long as you disable that extra camera

 

[mibc9394]

kinrean thanks man, solid advice. especially for the normal tutorial and hand swap and ref model is excellent planning the right topology and normals for the desired animation is something ill keep in mind do you have any good base models in mind for anime characters? i remember Guilty Gear XRD had some amazing looking anime style normals
View attachment 5537653

mibc9394
here you go bud may it be more useful to you then for me ^ ^

You must be registered to see the links

i used the comfyui manager to install missing files

I do have a culling system based on AABB camera frustum planes check, its not perfect the images on the side get popped in but it allows me to get good fps right now. the VAT method looks promising as well! Yeah streamable assets in unity is on my to do list for the level assets ^ ^ thanks so much for info
View attachment 5537660

btw using camera planes to check culling is actually quite expensive. I assume you are checking the six frustum planes? If that's the case, you need six checks for each instance. A better way is to do the checking in clip space. That requires one check per instance only

 

[kinrean]

kinrean thanks man, solid advice. especially for the normal tutorial and hand swap and ref model is excellent planning the right topology and normals for the desired animation is something ill keep in mind do you have any good base models in mind for anime characters? i remember Guilty Gear XRD had some amazing looking anime style normals


mibc9394
here you go bud may it be more useful to you then for me ^ ^

You must be registered to see the links

i used the comfyui manager to install missing files

I do have a culling system based on AABB camera frustum planes check, its not perfect the images on the side get popped in but it allows me to get good fps right now. the VAT method looks promising as well! Yeah streamable assets in unity is on my to do list for the level assets ^ ^ thanks so much for info

well, base mesh is simple just any model that you base on it or even mix 2 model together, you can even use guility gear model if you can find it ( i can't find good one).
as long as is close to nude, and is in a easy pose like t pose or a pose.

You must be registered to see the links

i not really high level for anime 3d, so i can't be too detail on it.
here some link i normal use.

You must be registered to see the links

You must be registered to see the links

https://f95zone.to/sam/latest_alpha/#/cat=assets/page=1/prefixes=42

if you asking for model for anime normal face.

You must be registered to see the links

You must be registered to see the links

You must be registered to see the links