Legacy Graphics Refactoring for Remasters
2015 project: textures 512×512, diffuse maps without metallic/roughness, meshes with 15,000 polygons for main character — all in Built-in Render Pipeline. Task: move to URP or HDRP, raise quality to modern PBR standards, don't break gameplay and don't increase system requirements beyond reason. This not "repaint textures" — systematic work with artistic production technical debt.
What breaks when changing render pipeline
Transition Built-in → URP — not automatic. Edit > Rendering > Render Pipeline Converter in Unity converts Standard materials to Lit (URP), but this baseline conversion. Custom shaders in GLSL or ShaderLab without Surface Shader API — need rewriting under HLSL or porting to ShaderGraph.
Main loss: Standard shader in Built-in supports Specular workflow (albedo + specular + gloss). URP Lit — only Metallic workflow (albedo + metallic + smoothness). Specular → metallic conversion not unambiguous: old materials often had physically incorrect specular values that gave "beautiful" result in old render and look wrong in PBR.
Specific situation: stone walls in 2015 game had specular color (0.2, 0.2, 0.2) with glossiness 0.3 — this gave soft gloss. After conversion metallic = 0.0, smoothness = 0.3 — walls look matte than should because F0 for dielectrics now computed physically. Needs manual smoothness tweak + Detail Normal Map addition for compensation.
Texture upscaling and reworking
Textures 512×512 in remaster for modern 1440p/4K monitors look bad. Options:
AI upscaling via Gigapixel AI or Stable Diffusion with img2img (controlnet tile for structure preservation). Gigapixel gives 4× without artifacts on tileable textures — 512 → 2048 in minutes. But for organic (character skin, plants) result unstable — needs manual refinement.
Full rework under PBR: artist creates new texture set from 4 maps (Albedo, Normal, Metallic/Roughness, AO) in Substance 3D Painter, using old texture as color reference. Right for key assets, unrealistic for whole game.
Normal maps from old geometry. If character has low-poly and high-poly versions (even in old quality) — rebake normal maps via Marmoset Toolbag or xNormal with correct cage offset. Old normal maps often baked without UV-seam separation — seams visible at edges, become more noticeable in new HDRP lighting.
Meshes: retopology vs LOD
Mesh rework for remaster depends on task. Game stays on same platform — aggressive retopology unnecessary. But if original for 7th gen consoles (PS3/Xbox 360 style), meshes often have non-functional topology: triangles everywhere, no edge flow for deformations. Adding new animations or Blend Shapes — facial animation on such topology impossible.
For remasters with updated cutscenes: retopology key characters, add blendshape-rig for facial animation, setup LOD Group with LOD0/LOD1/LOD2 — standard work.
LOD in URP/HDRP via LODGroup + SpeedTree-style billboard for far objects gives performance boost without visual degradation on near-LOD.
Shaders and visual effects
Old Particle System effects on Built-in Material — rewrite under VFX Graph (URP/HDRP). VFX Graph runs GPU, supports lit particles with normal maps and proper PBR lighting — quality jump for fire, smoke, magical effects.
ShaderGraph for custom shader rework: old 200-line Vertex-Fragment shader → visual graph with Sample Texture 2D, Normal Blend, Lerp, Fresnel Effect nodes. Debug and iterate easier, result compiles for all supported platforms.
Timelines
| Task | Timeline |
|---|---|
| Convert materials Built-in → URP (automation + fixes) | 1–2 weeks |
| Rework texture set for one character (PBR) | 3–7 days |
| Retopology + UV unwrap character | 5–10 days |
| Full project graphics refactoring mid-scale | 2–4 months |
Cost determined after asset volume audit and remaster target platform.





