Royalty retweeted
I love seeing how aggressive Epic is going after AI integration/ because itโs coming like a triple stacked freight train. Everyone should be running all ready! Godspeed! ๐ช๐ช๐ช
Replying to @Jvthewanderer @t_looman
What we are trying to do: Build an engine thatโs more powerful than UE5, easier to develop in than Unity, and is the most efficient effective way to build with AI acceleration.
The problem with having old systems and new systems side-by-side is the tendency to mix them, and then their interoperability becomes a giant tangle of blueprints and verse, actors and entities.
We thought that side-by-side was unavoidable, but AI assisted conversion is now clearly way more effective, and will produce a way more manageable UE6 project than a hybrid, while making it way easier for us to simplify and improve the engine.
Weโll be proving it out by converting Fortnite, the game with the largest base of blueprints in existence, to the new approach.
Royalty retweeted
Kept pushing animation further with one of my favorite materials.
M_Rect_Dissolve_V6 now has a dynamic rectangular burn with varying timing across the grid, plus layered geometric patterns.
Experimented with directional dissolve control and per-rect pulsing/offset animation.
Been busy lately so new materials are tough, but I still want to keep exploring.
Material Nodes: blueprintue.com/blueprint/rzโฆ
#UE5
Royalty retweeted
ใWIPใGlass Shattering Transition
Made a breaking glass transition with UE5 Post Process Material. The screen cracks and shards fly apart.
Still way too heavy and unoptimized, so itโs WIP. Been too busy to finish a polished version, but wanted to post while the idea was fresh.
Side note: UE6 has been announced with quite a few breaking changes. For now though, I'm just going to keep enjoying UE5 materials!
#UE5
Royalty retweeted
Replying to @games_inu
They won't. They'll just be not supported in UE6 unless the authors port them. Simple.
Royalty retweeted
We are seeking an Expert UE5 lighting specialist for a short-term remote consulting engagement to review our fully dynamic day-night lighting setup and help align the project's visual presentation with the Game Director's artistic vision. .@gtsplayground DM me!
Royalty retweeted
Kick-start your real-time VFX in Unreal Engine with the new Niagara Examples Pack๐
50+ VFX systems you can explore, break apart, and adapt to your needs, including explosions, ballistic impacts, sparks, fire, and more. โจ๐ฅ
Out now on @Fab!
epic.gm/niagara-samples
Royalty retweeted
Screen Shot Saturday dont let your dreams be non-voxelized nanite static meshes because you can do that now in Unreal Engine 5.7!
Royalty retweeted
Bookmark this comprehensive video course on rigging in Unreal Engine. This is taught by the team that builds Unreal's rigging tools.
dev.epicgames.com/community/โฆ
Royalty retweeted
We gonna gain another 10 fps with dis right here .@gtsplayground
UE5ใฎใตใใฌใใซใฎ่ชญใฟ่พผใฟๆใฎใใใๅฏพ็ญใฎๆนๆณใใใใชใใฌใใใช่งฃ่ชฌใใฆใใ่จไบใๆ็จฟใใใพใใใ
dev.epicgames.com/community/โฆ
ๆ้ฉๅใฎๅญฆ็ฟใใใ้ใฏใ็ขบ่ชใใฆใใใๆนใ่ฏใใใใช่จไบใงใใ
Royalty retweeted
Replying to @RoyaltyGameDev
Thanks for the tip!
Using the drag I was able to finally get a stable system. Here's how it looks if you're curious:
teapawt.org/roguephotondev/status/โฆ
Improved Tornado AoE!
The physics are much more stable now: units stay in the vortex, spiral upward, then fall back naturally once they escape the forces.
Feels much closer to what I had in mind.
Tested both on friendly units and mid-combat ๐
#ScreenshotSaturday #indiedev #RTS
Royalty retweeted
Replying to @RoyaltyGameDev
I pass in velocity, hit on collision life adv amount, lifespan, and then pass out dealths as impacts :) mesh distance fields for particle collisions, no tracing teapawt.org/gamedevmicah/status/19โฆ
Just a little basic basic fire monster test... just trying some grayboxing out. I like it, I want to cut down some buildings with a beam... .@Godzilla_Toho Godzilla Shinzilla Unreal Engine 5 UE5 Gamedev Baby! LET ME IN!!๐ฅ๐ฅ๐ฅ
Royalty retweeted
Replying to @RoyaltyGameDev
Is this the same as pooling the projectiles?
I tried making a pooled projectile system in C++, but I got around 7ms with 1800 projectiles (tick is also pooled into a central thinker, but still uses capsule colliders and no ISM)
teapawt.org/MakotoIchinose6/statusโฆ
Bullet hell test in #UnrealEngine, coded from scratch.
Collision has to work so Niagara only is a no go, and movement component stacks up so bad. So I pre-pooled projectiles and have Combat Manager Subsystem move them all at once.
1800 moving bullets takes up ~7ms
#gamedev
Royalty retweeted
At some point it's not about pushing big amounts, but pushing less for the same effects. The new UE foliage rendering for example starts replacing geometry with voxels at a distance. Worth considering.
Royalty retweeted
Replying to @GameDevMicah @RoyaltyGameDev
When I reach a point where this can be shared, it's currently a mess of scratch code. I'll tidy it up and push it to GitHub for you both to review and critique my coding skills. :D
Be gentle.. and no.. I won't apologise for magic numbers.. you're not my coding jebus.๐คฃ
Royalty retweeted
Replying to @RoyaltyGameDev
I'll try and answer via twitter best i can :D
I tried the add remove clear approach, but it just churns memory and rebuilds HISM clusters for little gain. I now preallocate a global pool per mesh type so capacity is reserved up front. Each instance has a stable slot. Adds write a transform into a free slot. Removes mark the slot inactive and schedule a batched transform to park it out of view or scale to zero. HISM does not have a true per instance visibility flag, so even hide and show is a transform update. The cost is controlled by batching and by touching only the affected clusters.
When demand exceeds the main pool I spill into a shadow pool for immediate visuals, then migrate those instances back to the main pool over several frames under a strict budget. Worker threads only build the transform arrays and index lists. The actual HISM calls run on the game thread in batches. Spikes like many explosions can stress timing, so the drain rate is capped and backpressure is applied when budgets are exhausted.
The model is drain and expand. Visibility gives headroom because this is a top down camera. I use a simple cluster grid sized to the edit brush and view. Shapes mark overlapping clusters in constant time. I then touch only the handles inside those clusters and flush one batch per changed target. An FPS would need larger budgets but the approach holds (Horizon cameras can be costly for static mesh art pipeline vs pure cubic voxels sadly).
Instance storage and generation are separate from rendering. View tests use sparse sets, flood fill, and bit packed grids on non critical frames with a small token budget so only a few tiles update each frame. I am experimenting with ray cast helpers, but the expectation is that flood fill or a small distance field will give better cost stability.
Collision is minimal. NPCs carry a small local nav mesh or use nav invokers. Colliders are toggled by proximity or replaced with simple proxies. I avoid per instance physics on HISM. If physics is required, bodies are pooled separately and synced at cluster granularity off the render path. Chaos can play a role here per @GameDevMicah explosion mad man work ๐คฃ but yet to get to that point where I can see through his lens on the how.
The predefined geometry used for culling and boolean edits lets me predetermine collision outcomes in the gameplay layer. HISM only receives batched adds, batched transform updates, and batched removals that reflect those decisions. The pool owns all state and indices. Overflow exists to absorb bursts and drains back under budget without exceeding frame time.
You can then use RLE compression and other algos to help balance out needless loops. Transfering between CPU and GPU timing is also a bottleneck I have to step back and think on more.
I should mention that pools are fixed sizes as well to fit into L1 so there is overallocation in memory at times where not needed. The goal is fewer cache misses and faster tight loops
Royalty retweeted
Replying to @GameDevMicah @RoyaltyGameDev
Side note: I'm also cheating with @claudeai here by getting it to parse profile(s) to help me isolate failure(s) or spikes or perf issues to compare with the code I wrote.
It pretty much confirmed my suspicion my draw calls were bloated. ๐ Like I just made a bad change just then in code and it helped take the csv profiling and compare to which line of code is shitting the bed.
I wouldn't let it take the wheel though on HSIM. I tried that once.. didn't end well.






















