Hi all, we are building a character like Siren, the entire scene, has NO background, just black and the character itself.
Models are pushed to high poly (about 250 to 500k poly I think), textures all 4ks, with albedo, roughness, spec, AO, norm, metallic and subsurface profile. Culling distance disabled. Textures set to ‘character’ and filter sharpen 3-5 (hoping it’ll help). I also set all the cinematic/ inset shadows (or something like that which made the char looked a lot better in terms of lighting.)
Overall lighting I set to 0.5 quality 10 bounces 10 bounces 5 quality (sorry forgot what were they named, but it helps push stuffs up too for light)
But graphic still looked not very high quality… what other things should I look into?
Take note I am a self taught UE artist that runs a small game dev co (if u guys do see the posts I always do, looking for artists), so do bear with me for some of the technical terms that I might not be putting it at correctly.
Skeletal meshes don’t light the same as static ones no matter what.
if all you need is a Beauty Shot you should export the posed mesh as static.
Also, the engine could be pushing different mips than what you expect. It’s optimized to perform, not to look good. Either bias them or disable mips altogether in the material texture sample.
If you need to animate it and shoot a short. Cranking things up is OK. It works in a pinch. Render times can take a day or 2 for a full scene though.
you can still get really good results, but you have to really fine-tune the directional light or only use fine-tuned spot/flood/rect lights.
Also, the scheletal mesh light settings needs to be setup right.
Consider also building the engine from source and bumping down the per vertex influence settings cut off so that the animations of the short act much closer to what you get inside a DCC (I think the default is .01, changing it to .001 would give you much more accurate skinned meshes).
Definitely enable the engine settings for skinned cache and adjust all of the relevant materials for the characters to use a derived tangent basis. It’s just better regardless of the limitations/issues it could bring.
If you need this to work in a game… then you need to bite the bullet and settle with a level of quality that doesn’t compromise performance. But that’s a whole topic of its own anyway.
It’s not for a game but for something like Siren level. It’s real time character which animates and talks based on random set of animation we programmed. Need to know how Siren achieve that level of real time quality. Might be something we missed along the way.
Let me rephrase.
if its live, you have to optimize for performance.
If its just filmed, you do not.
Also, it’s hard to tell what you mean by it not looking good without seeing it and you haven’t described what it is that doesn’t look good.