Chapter 3: Lighting Renders for Paintover & Material IDs

Created by Sarah Choi (prompt writer using ChatGPT)

Lighting Renders for Paintover & Material IDs

2D / 3D Hybrid Methods for Character Concept Artists (Blockouts, Sculpt Bases, Scans, Photobash Ethics)

Lighting renders and material ID passes are the glue between 3D truth and 2D speed. A disciplined light rig reveals planes, preserves silhouette, and compresses values into clean, editable blocks. Material IDs then turn those blocks into selections you can grade, repaint, or relight non‑destructively—without inventing pixels that won’t survive in production. Whether you are on the exploration side doing rapid paintovers or on the production side preparing references for modeling, surfacing, and lookdev, a reliable render+ID workflow reduces rework and keeps everyone aligned to the same form and material logic.

Begin from the approved blockout or sculpt base, not from a beauty pose. Lock camera height, focal length, and distance to match the shipping view (or the marketing lens you intend to use), and stamp the camera settings on export. Consistency makes comparisons honest across ideation, iteration, finals, and handoff. Neutralize scans immediately if they are present: strip baked lighting, assign a flat clay material, and ensure the mesh sits in engine units. You want lighting to describe construction, not photography to fake it.

A studio‑neutral rig is your workhorse. Set a frontal key that defines the dominant planes, a broad fill that lifts midtones without erasing edge flow, and a rim that separates the silhouette from neutral gray. Work in linear color; meter exposure for the mid‑gray of your clay so highlights don’t clip and shadows don’t crush. Avoid drama at this stage—no colored gels, no stylized gradients—because extreme lighting disguises design problems and trains marketing to expect moods the game might not deliver. If the project will ship with a strong HDRI look, add a second “PBR validation” preset that uses a simple environment light plus a small kicker; it should confirm that value hierarchy still holds under realistic energy distribution.

Render a minimal but powerful set of passes for paintover. Besides the beauty (or clay) image, export ambient occlusion (broad + tight if available), normals (world or tangent), curvature (concave and convex), shadow‑only, specular‑only, and depth (Z). AO and curvature provide instant material suggestions (dust vs. polish), normals support quick relighting in 2D, shadow‑only helps you paint contact and weight, and depth allows atmospheric separation and focus pulls for marketing tests. Keep denoisers conservative; over‑smoothed AO and specular passes produce plastic results when graded.

Material IDs turn parts into selections. Create an ID pass in which each material class receives a unique, high‑saturation hue at full value. Group IDs by shader behavior, not by color fashion: “painted metal,” “bare metal,” “leather,” “cloth—stiff,” “cloth—drape,” “skin,” “hair,” “glass,” “emissive,” “FX proxy,” and so on. Keep a master palette and reuse hues consistently across characters so studio actions and LUTs remain portable. Render IDs with hard edges and minimal anti‑aliasing to preserve clean selections; you can feather in 2D when needed. For layered items (e.g., a leather strap over armor) ensure the top layer’s ID masks fully occlude what sits beneath to prevent selection leaks.

In 2D, build a non‑destructive paintover stack that respects the render. Start with the clay beauty at the bottom. Above it, place AO as multiply at low opacity, shadow‑only as multiply with a mask that you paint into for contact decisions, specular‑only as screen or linear dodge to test material sheen, curvature as a subtle overlay for edge pick‑outs, and a normals‑driven relight layer if your tool supports it. Keep a folder of color correction layers clipped to each material ID selection: this is where you push hue, value, and roughness impressions without repainting the world. If a photobash element enters the stack, compel it to obey your lighting by matching perspective, focal length, and value range, then feeding it through the same AO/shadow/spec layers so it participates in the same physics.

Photobash ethics and provenance matter. Use licensed sources or studio packs, store raw images in a “Sources” group with IDs, and transform sufficiently that no single source defines a hero read. Never bash from another concept artist’s stylized page; you import not only pixels but their design language. If a photographic edge contradicts your material ID boundaries (for example, a jacket photo bleeding into a pauldron zone), fix the selection and repaint the transition so production inherits clear intent. The render and its IDs are a contract for construction, not a collage of exceptions.

For hair, skin, and FX proxies, use IDs to communicate behavior clearly. A single “skin” ID is insufficient if face and hands require different shader responses; separate IDs for “face—micro‑spec” and “hand—callus” help lookdev plan sliders and masks. Hair benefits from two IDs—“mass” and “edge flyaways”—so paintovers can test rim control and spec breakup. FX anchors get their own ID and sit on a separate pass (or layer) so UI/VFX can place sockets and test glows without polluting the clay values. Emissives should be painted as masks first, then graded with additive layers; otherwise they bake a mood into the base that may not survive engine calibration.

On the production side, material IDs accelerate surfacing and LOD policy. When IDs map to shader families, surfacing can author master materials once and apply them broadly. Include a short legend that states roughness/metalness intent per ID (e.g., “painted metal: mid roughness with edge wear; bare metal: lower roughness, higher spec energy; leather: anisotropic gloss on pull zones; cloth—drape: high roughness, AO‑heavy folds”). For LOD planning, note which IDs merge or vanish at distance (small buckles collapse into “metal—generic”; stitching texture disappears into “cloth—drape”). Rigging also benefits: IDs that align with deformation zones tell riggers where to expect sliding plates vs. skin‑bound cloth.

Keep exports clean and portable. Use 16‑bit PNG or EXR for beauty and depth to preserve grading latitude, 8‑bit PNG for AO, normals, curvature, and IDs. Name files predictably: CHAR_codename_CAM35mm_V03_beauty.png, …_AO_wide.png, …_normals_world.png, …_ID.png, etc. Stamp a small legend on the ID page mapping hues to material names, and embed a text file listing camera and exposure settings. A simple checklist at export—camera stamp visible, ground plane and scale bar on, linear color confirmed—prevents the expensive kind of “why does this look different?” email.

Common failure modes are avoidable with discipline. Over‑stylized lighting erases plane truth and makes paintovers lie; fix by returning to the studio‑neutral rig before you add flavor. Inconsistent focal length makes lineups incomparable; lock focal and distance per character class. Denoiser sheen and crushed blacks hide cloth logic; keep ranges wide and verify at gameplay scale. ID collisions (reusing the same hue for two materials) or fuzzy anti‑aliased edges make selections brittle; maintain a palette and render hard masks. Finally, letting photobash drive the value structure trains everyone to a look the engine cannot reproduce; always grade photographs to the render, never the other way around.

When executed well, lighting renders and material IDs do quadruple duty. They give concept artists fast, truthful paintovers; they give surfacing and rigging actionable maps; they give marketing honest crops that the game can match; and they give the whole team a shared, testable picture of the character under light. The result is a pipeline where taste lives on top of evidence: the shape language proven in blockouts, the physics checked in sculpt bases, and the materials and moods you paint today can ship tomorrow without surprises.