- Mar 22, 2019
- 593
- 361
Soon I think, the new version 11 is now available on Patreon, so maybe in a week or two
Unreal Engine Delta Zone [v20.2] [DEVOLUTION]
- Sep 28, 2020
- 189
- 1,156
on each of these things I have already spent months, it's just a basic stuff that everyone does, to get more performance I need to go much deeper inside, which takes lots of effort and cuts down from actual game features development which I need to balance
- Sep 28, 2020
- 189
- 1,156
I have been kinda lazy to write detailed update notes but in general
- whole new map (city outskirts)
- new cyberware function with disassembling hand, different holding of gun (can hold without handguard etc.. and recoil changes)
- new driveable hoverbike with custom coded physics and crash physics
- tons of bugfixes from new features (bunch of them still remain but its not too bad)
- tons of small changes, polish that is hard to track
- Sep 28, 2020
- 189
- 1,156
I eventually will as always
- Sep 28, 2020
- 189
- 1,156
related to new nsfw content, to be transparent, I hadn't had that much luck finding new animator who is capable reaching quality, animators are really hard to come by, most of them have work ethics below zero or are already hired in aaa studios or just don't want to animate nsfw stuff.
anyway if anybody knows about someone suiting this project I would gladly get someone onboard
this month I will try again to get someone reliable to help out on it so each release has at least two new h-animation types (4 loops per type)
otherwise the project is going pretty well, step by step I'm turning all those ideas and plans into reality
anyway if anybody knows about someone suiting this project I would gladly get someone onboard
this month I will try again to get someone reliable to help out on it so each release has at least two new h-animation types (4 loops per type)
otherwise the project is going pretty well, step by step I'm turning all those ideas and plans into reality
- Mar 22, 2019
- 593
- 361
How hard is it to learn how to do NSFW animation? When I search a little on the internet, there is practically no thing interesting enough to initiate to make it..
- Dec 1, 2017
- 269
- 479
Depends. I used to make content for SecondLife. Making objects in Blender, no problem. Texturing them, no problem, scripting them, no problem. Animating, NOPE... I dipped into animation and couoldn't wrap my head around how time consuming it is. Every damn frame has to be adjusted over and over. Now, I'm going in with no one teaching me, so I likely was doing everything wrong, but I self taught myself everything else too and that one I just couldn't do. Now.. I bet if you had Motion cap hardware, it be 10x easier and quicker.
Edit: and yeah, you really can tell the difference between mocap and hand animation.
- Sep 28, 2020
- 189
- 1,156
it absolutely doesnt matter if its normal or nsfw animation, skill required is same
Its easy to find resources, but it takes lots of artistic skill and feel for the movement
- Aug 4, 2020
- 2,525
- 3,208
Did kind of feel off about "I've already squeesed every last fps out of it" when you said it but I shut up because I didn't want to be a jerk, but this does sound more like the reality.
Agree you need to balance your work but saying that you squeesed every last fps out of it, when the game runs pretty meh with not much going on, is a bit much.
Thanks for the clarification tho.
- Sep 28, 2020
- 189
- 1,156
compared to my time budget I spent already way too much time on optimization and I'm running out of optimization opportunities atm
also compared to other similar-looking games game runs pretty good, don't have too much complaints about performance and also you need to consider other studios have at least 3-5 full time experienced devs doing nothing all time except trying to find way to get better performance without losing too much visuals. and even with that, people complain about ubisoft doing downgrades etc..
to be honest it's a miracle those games run in such performance, and people have no clue, call them out for "not optimizing tames" even tho so much insane talent and manhours went into it
D-E-V-O-L-U-T-I-O-N = Alright, so you got all the basics covered. So some further ideas for optimization:
1] How high is the polygon count / texture resolution on each Tentacle. You might be able to optimize them when they aren't visible or at a distance away.
2] When a tentacle travels through the water, is the water just a textured surface, or is it an actual mesh that deforms as the tentacle passes through it?
3] I know the game pretty heavily relies on fog. If you totally disabled the fog, does it improve performance? If so, then try to look into optimizing the fog in some fashion.
4] How does the game handle darkness? Is it merely the absence of a light source or is it using soft-shadows to instead?
5] When the AI is not directly seeking out the player, how expensive is their pathfinding? If it is quite expensive, then you might consider patrol routes instead of them wandering randomly. That way you could also decide their specific path ahead of time.
6] Are you performing physic calls on the grass, trees, shrubs, etc... Witcher 3 had a very well-optimized way of doing that (mostly by disabling the collision on the plants, as they'd clip right through walls when it was windy)?
7] When you have weather effects, are they particles, sprites, etc...? Like rain drops...
8] Since the later builds allows for visual clothing customization, are all the meshes that can potentially be worn, pre-rendered and thus taking up valuable resources (ie RAM)? I ask because Witcher 3 would pre-render everything in your inventory, so you could quickly view the mesh of any object?
9] How high are the polygon of non-essential map elements? Rocks, fences, trees, etc...? Services like SpeedTree might optimize their meshes, without too much loss of quality.
10] Same idea as 8, but now for the weapon parts? Or items in your inventory.
11] Since you can have belly expansion with moving tentacles and bulging of eggs, do these objects have a high polygon count / textures of their own?
12] For shooting mechanics, is it using Tracers or simulated bullets + trajectories?
13] When an enemy detects the player, do they always know where you are, or do they have to maintain visual conal lock?
14] How often do you use double-sided meshes? These are like when you have an armor vest, that has the outer section and the inner layer rendered; most games use double-sided meshes sparingly, as appropriate.
15] When you are in 1st person, is the 3rd person mesh fully rendered? Similarly, in 3rd person, is the 1st person mesh fully rendered?
16] What does the enemy "see" when they shoot at you? Is it a fully rendered object, or a simple object, with a low polygon count?
17] Is all the items in a chests, looted or otherwise, rendered meshes, or only rendered when removed from the chest?
Probably have more ideas later.
1] How high is the polygon count / texture resolution on each Tentacle. You might be able to optimize them when they aren't visible or at a distance away.
2] When a tentacle travels through the water, is the water just a textured surface, or is it an actual mesh that deforms as the tentacle passes through it?
3] I know the game pretty heavily relies on fog. If you totally disabled the fog, does it improve performance? If so, then try to look into optimizing the fog in some fashion.
4] How does the game handle darkness? Is it merely the absence of a light source or is it using soft-shadows to instead?
5] When the AI is not directly seeking out the player, how expensive is their pathfinding? If it is quite expensive, then you might consider patrol routes instead of them wandering randomly. That way you could also decide their specific path ahead of time.
6] Are you performing physic calls on the grass, trees, shrubs, etc... Witcher 3 had a very well-optimized way of doing that (mostly by disabling the collision on the plants, as they'd clip right through walls when it was windy)?
7] When you have weather effects, are they particles, sprites, etc...? Like rain drops...
8] Since the later builds allows for visual clothing customization, are all the meshes that can potentially be worn, pre-rendered and thus taking up valuable resources (ie RAM)? I ask because Witcher 3 would pre-render everything in your inventory, so you could quickly view the mesh of any object?
9] How high are the polygon of non-essential map elements? Rocks, fences, trees, etc...? Services like SpeedTree might optimize their meshes, without too much loss of quality.
10] Same idea as 8, but now for the weapon parts? Or items in your inventory.
11] Since you can have belly expansion with moving tentacles and bulging of eggs, do these objects have a high polygon count / textures of their own?
12] For shooting mechanics, is it using Tracers or simulated bullets + trajectories?
13] When an enemy detects the player, do they always know where you are, or do they have to maintain visual conal lock?
14] How often do you use double-sided meshes? These are like when you have an armor vest, that has the outer section and the inner layer rendered; most games use double-sided meshes sparingly, as appropriate.
15] When you are in 1st person, is the 3rd person mesh fully rendered? Similarly, in 3rd person, is the 1st person mesh fully rendered?
16] What does the enemy "see" when they shoot at you? Is it a fully rendered object, or a simple object, with a low polygon count?
17] Is all the items in a chests, looted or otherwise, rendered meshes, or only rendered when removed from the chest?
Probably have more ideas later.
- Mar 22, 2019
- 593
- 361
Ohw fuck...
Now I understand why there are very few 3D games with very good animation..
I see, thank you for your answers
I'm interested to see and understand how this kind of thing is done, since in general, people who show how to do it, only do SFW, and that in school, daring to talk about it is literally taboo... that intrigued me a lot
Well its a lot like programming. Many people can code. Only some can code very elegantly, efficiently, and create wondrously dynamic systems.
- Sep 28, 2020
- 189
- 1,156
1. automatic mesh density set byhow much percentage of pixels of screen cover, and biased by game graphics settings
2. when tentacle is underwater mesh is disablet to get perf. ripples are just 2d textures on top (ditched runtime virtual textures to get more performance as it was not worth it)
3. fog cant be turned off as it would drastically break gameplay,I heavily optimize fog and change its depth slices, z grid resolution and 5 other variables based on settings to get optimal fps
4. interior darkness is my inhouse lighting solution that has good performnace (compared to lumen in ue5)
for soft shadows in cloudy weather I just downscale resolution to get better look + performance gets a bit better.
ten I use a lot of screen space shadows for tiny objects with sunlight and also disable shadows on most of enviro lights to get tons of performance back so they only cast screen space shadows
5. AI cpu code optimisation is super heavy it would take hours to explain properly. but if you dont look directly at them or they are behind the wall, I instantly downscale frequency of their brain, downscale animation updates, also remove customizable weapon. really heavy optimization, also when they are too far, they turn into proxy characters which is completely separate optimization tech
6. I have very little physics in game atm except for hoverbike when you crash and enemy ragdoll, and custom hoverbike physics code which runs pretty good
6. vegetation has zero physics, all wind is handled with shaders in gpu
7. rain is tons of particles, but has almost zero impact on performance when I tested it
8. I switch textures of clothes on runtime, to not kill memory completely, they have virtual texturing so there will be always some delay for loading hi-res ones
9. resolution of these meshes doesn't really matter (except memory but polygons are ok compared to huge textures for memory) it doesn't matter because its entirely screen dependent (far away objects will never reach maximum detail becuase of lodding (I set up how high poly count relative to screen size should there be either automatically or manually if needed)
10. I try to make sure guns have max lod in first person and then once far away I aggressively reduce them to about 5% of original poly density
11. I use realtime tesselation on 3rd person character so its like infinite resolution as much as you zoom in. Bulges are handled by baked displacement and normal texture flipbooks (tesselation doesn't really cost that much for how much it improves detail)
12. for bullet trajectory I use both raycasting and physical projectile that has real life behavior (air resistance, bullet drop in distance, velocities)
13. by code, AI never gets position of player. they only get their own object that acts as their memory of enemy, if it was sound, sight or laser/flashlight beam, then direction and speed of last seen movement so they can try predict
there is also separate ai object that raytraces and scans for line of sight (it accurately monitors density of vegetation, or if player peeks through tiny hole or there is tiny object covering player (so thing like can AI vision block in half life 2 doesnt happen) AI is incredibly complex to explain it would again take few days to explain properly )
14. by default I never use it, only when its necessary also collisions arent double sided
15. its by far biggest optimization that in each view mode other mesh is fully disabled and invisible to get lots of performance back
16. I had a plan for approximation mesh with body hands legs etc... atm I just use sphere that is scaled and moved to closely simulate what player does and its more than good enough atm. (approximation of enemy seeing you arms and shoulder while you cant see them with head behindthe cover works well)
only thing that is fake or wrong is they don't really see your legs which is a direct player advantage to remove frustration
17. in chests there are no items or nothing is rendered. it gets generated when chest is opened for first time and is just a struct (bunch of variables with all item stats) this has zero gpu performance and some cpu cost on generation which is nothing
atm my biggest trouble is texture res/vram optimization, using virtual textures to help it, and overall memory management, in lastest 11 release it seems to be working well finally sometimes I check profiler to see if anything big pops up that I can optimize but usually when I code it I already optimize things on first go.
each answer here is extremely simplified and would take hours to answer properly and on each optimization thing I spent weeks/months already hope it at least looks like I have very good idea what I'm doing
Last edited:
Alright thanks for the answers, D-E-V-O-L-U-T-I-O-N, so some other ideas for you to ponder on:
1. Can you reduce the Post-Processing on objects at higher draw distances? Like say, everything within 25 meters get full post-processing of say 8x AA, then within 26-40 meters AA and other features are reduced by half (so AA at 4x), then 41-55 meters, halved again (or a quarter of post processing), then beyond that, no post-processing? Or is that what UE4 does automatically?
2. Can you use a similar technology to John Carmack's Mega Textures, implemented in IDTECH 5/6 (Rage / Wolfenstein The New Order)? So instead of several individual textures, you load in one massive texture that handles everything in the level? Or is that how you do Virtual Textures?
3. Does Occlusion Culling work automatically in UE4, or do you have to create a method for yourself? In that, OC would prevent the mesh from being rendered if the entirety of the mesh wasn't in your viewport or would reduce the amount of rendered mesh, based on how much was in your viewport.
4. Is there any way to do a "batch" of post-processing in a single draw call? ie Do AA, Antistrophic Filtering, and Bloom, all in one draw call? Having too many draw calls (usually thats 10,000+ per frame for reference) can seriously kill the framerate.
5. So from your tests, what is the performance bottleneck for the game? Is it the RAM, VRAM, GPU Core Speed, CPU Core Speed, or Hard Drive Access speed? And what are the typical numbers for consumption on your rig, when running on Ultra (and framerate)?
6. Does the game suffer from framerate pacing issues at all? That could be part of the problem.
Anyway, I'm going out, so if I can think of something else I'll mention it.
1. Can you reduce the Post-Processing on objects at higher draw distances? Like say, everything within 25 meters get full post-processing of say 8x AA, then within 26-40 meters AA and other features are reduced by half (so AA at 4x), then 41-55 meters, halved again (or a quarter of post processing), then beyond that, no post-processing? Or is that what UE4 does automatically?
2. Can you use a similar technology to John Carmack's Mega Textures, implemented in IDTECH 5/6 (Rage / Wolfenstein The New Order)? So instead of several individual textures, you load in one massive texture that handles everything in the level? Or is that how you do Virtual Textures?
3. Does Occlusion Culling work automatically in UE4, or do you have to create a method for yourself? In that, OC would prevent the mesh from being rendered if the entirety of the mesh wasn't in your viewport or would reduce the amount of rendered mesh, based on how much was in your viewport.
4. Is there any way to do a "batch" of post-processing in a single draw call? ie Do AA, Antistrophic Filtering, and Bloom, all in one draw call? Having too many draw calls (usually thats 10,000+ per frame for reference) can seriously kill the framerate.
5. So from your tests, what is the performance bottleneck for the game? Is it the RAM, VRAM, GPU Core Speed, CPU Core Speed, or Hard Drive Access speed? And what are the typical numbers for consumption on your rig, when running on Ultra (and framerate)?
6. Does the game suffer from framerate pacing issues at all? That could be part of the problem.
Anyway, I'm going out, so if I can think of something else I'll mention it.
- Sep 28, 2020
- 189
- 1,156
1. aa works very very differently, I use TAA g4 with upscaling similar to dlss
2. I already use it its called virtual texturing
3. occlusion culling works and its super aggresive
4. aa and deffered rendering is inbuilt in engine and probably as optimized as it gets
5. atm its shader complexity, resolution, fog resolution, amount of objects that cast shadows and pretty much gpu things vram memory if fulli filled is issue too but with virt tex its much better
1080ti should run 2k on ultra on 60fps
6. if you look at sky there is nothing to calculate so fps is better, if you look at transparent water you need to calculate 2x more so fps is different again
1. Offering other non-intensive forms of AA can help. TXAA is one of the newest types of AA, so, it doesn't work as great on older hardware. Same idea with other forms of post-processing. You could have it set in compatibility settings, to change various intensive post processing methods to lower-cost ones, for low-end machines (my machine is technically low end at this point, even though in 2015, it was top of the line).
2. Okay.
3. Okay.
4. Okay.
5. Well, then:
-> Many games have gotten away with lower quality shadow maps, of 1k size. Original skyrim did that for most of its shadow maps, and people rarely noticed bad shadowing; except closeups on characters. For things that need more detail, 2k is usually sufficient, and 4k+ is overkill.
-> Try to avoid upscaling resolutions. Any game that has used upscaling to reach a higher resolution, usually has to render the same screen 4 times and then combine the image together. Like many initial games that said their game was 4k supported, was 960p x4 screens, merged together to get the 4k effect, which has a much higher performance cost than just having the 4k screen rendered from the beginning.
-> With fog resolution, can you have two volumes/layers of fog? So the leading edge of the fog could be a higher resolution, but the rest of the fog is lower resolution? OR... have a volume centered around the player and the base volume of fog, and only the volume around the player is higher resolution. If that makes any sense.
-> Can any of those objects that cast shadows be set to static shadows over dynamic shadows? UE4 as you probably know is super efficient with baked shadows, but it has issues with dynamic shadows and thus suggests limiting them as much as possible. Additionally, can any of those objects be set to not cast soft shadows, as soft shadows is quite expensive.
-> Additionally, Skyrim's settings file had quite a few settings for manually modifying shadows, and maybe some of that information might be useful to you, for fine-tuning your shadows costs.
-> Can you reduce the number of objects that cast shadows? ie instead of having 5 trees next to eachother casting individual shadows, replace with a very large tree casting 1 shadow. Or having grass placed intelligently under a tree's shadow so that the grass doesn't have to cast a shadow.
-> For the GPU things, can any of it be sent to regular RAM instead? Most users have more RAM than VRAM and their RAM is often better cooled than their GPU VRAM. Also does your method of virtual textures abuse the VRAM or uses RAM? I believe if memory serves, that ID TECH shoved the data into RAM, and then extracted specific things from there into VRAM on a need-to-use basis, and then unloaded it once it was no longer needed.
6. I can't comment on other users, but I often will manually degrade water quality, to get better quality elsewhere. If the water is murky, I try to make it clear and visible; I like to see where I am going. So you might be able to reduce water quality without users noticing. I certainly don't notice lower water quality, unless its a game where I'm underwater a lot... lot Subnautica.
Hope some of that helps.
2. Okay.
3. Okay.
4. Okay.
5. Well, then:
-> Many games have gotten away with lower quality shadow maps, of 1k size. Original skyrim did that for most of its shadow maps, and people rarely noticed bad shadowing; except closeups on characters. For things that need more detail, 2k is usually sufficient, and 4k+ is overkill.
-> Try to avoid upscaling resolutions. Any game that has used upscaling to reach a higher resolution, usually has to render the same screen 4 times and then combine the image together. Like many initial games that said their game was 4k supported, was 960p x4 screens, merged together to get the 4k effect, which has a much higher performance cost than just having the 4k screen rendered from the beginning.
-> With fog resolution, can you have two volumes/layers of fog? So the leading edge of the fog could be a higher resolution, but the rest of the fog is lower resolution? OR... have a volume centered around the player and the base volume of fog, and only the volume around the player is higher resolution. If that makes any sense.
-> Can any of those objects that cast shadows be set to static shadows over dynamic shadows? UE4 as you probably know is super efficient with baked shadows, but it has issues with dynamic shadows and thus suggests limiting them as much as possible. Additionally, can any of those objects be set to not cast soft shadows, as soft shadows is quite expensive.
-> Additionally, Skyrim's settings file had quite a few settings for manually modifying shadows, and maybe some of that information might be useful to you, for fine-tuning your shadows costs.
-> Can you reduce the number of objects that cast shadows? ie instead of having 5 trees next to eachother casting individual shadows, replace with a very large tree casting 1 shadow. Or having grass placed intelligently under a tree's shadow so that the grass doesn't have to cast a shadow.
-> For the GPU things, can any of it be sent to regular RAM instead? Most users have more RAM than VRAM and their RAM is often better cooled than their GPU VRAM. Also does your method of virtual textures abuse the VRAM or uses RAM? I believe if memory serves, that ID TECH shoved the data into RAM, and then extracted specific things from there into VRAM on a need-to-use basis, and then unloaded it once it was no longer needed.
6. I can't comment on other users, but I often will manually degrade water quality, to get better quality elsewhere. If the water is murky, I try to make it clear and visible; I like to see where I am going. So you might be able to reduce water quality without users noticing. I certainly don't notice lower water quality, unless its a game where I'm underwater a lot... lot Subnautica.
Hope some of that helps.