Color grading 3D architectural renders in DaVinci Resolve

A render straight out of Cycles or V-Ray is technically correct and emotionally flat. The job of post is to take that technically correct image and give it the look of a photograph taken by a person with intent. For me, that work happens in DaVinci Resolve, not Photoshop, and I'll explain why.

This is the post-production process I use on most interior and exterior renders. It's not the only valid workflow. It is one that works fast and reproduces consistently across projects.

Why DaVinci, not Photoshop

Photoshop is the default in archviz. I switched to DaVinci four years ago for three reasons:

1. Node-based grading. You can re-order the look, isolate parts of the image, and non-destructively undo. Photoshop's adjustment layers are clumsy by comparison.
2. Better color space handling. DaVinci natively understands ACES and other color pipelines. If you're rendering in a wide-gamut color space (which you should be), DaVinci doesn't fight you.
3. Identical workflow for stills and animations. When a project needs a walkthrough later, my color pipeline is already sitting in DaVinci. No tool switch.

The cost: DaVinci is overkill for one-off touchups. If you only do five renders a year, Photoshop is fine. If you ship two a week, DaVinci pays for itself in saved time within a month.

The base node tree

Every render goes through the same starting tree. I save it as a powergrade and apply it to every shot.

That's the spine. Individual shots get extra nodes for specific problems (a too-bright lamp, a wall that came out too saturated, etc.) but the spine handles 80% of the work.

The grading moves that matter for archviz

I'll talk about the moves rather than the technical settings, because the settings depend on the source image.

Lift the shadows just enough.

Architectural renders tend to have crushed shadow detail because lighting is often set up for the hero areas (windows, focal walls). Lifting the shadows by 5 to 10% reveals texture in the corners and makes the room feel inhabited. Don't overdo it. Crushed shadows are still better than washed-out ones.

Warm the midtones, cool the highlights.

This is the cinematic look that defines premium photography: warm where the eye lingers (skin, wood, fabric, lamplight) and cool where the air moves (sky, glass, distant detail). Even a daylight render benefits from a small move in this direction. The eye reads it as "real photo, taken by someone with taste."

Reduce color saturation by 5 to 10% globally.

Renders out of Cycles or V-Ray tend to over-saturate by archviz standards. A slight global desaturation gives you headroom to push specific colors in localized nodes (the warm window glow, the brand-color throw pillow) without the whole image looking cartoonish.

Isolate the highlight glow.

If your render has windows or warm lamps, use an HSL qualifier to select just the warm bright pixels, then push saturation up 10% on that selection only. The image gets a sense of "lived-in warmth" without affecting the rest of the grade.

Cool the daylight bounce.

If you have a window with sky bounce coming in (you should), use another HSL qualifier to grab the cool blue ambient and reduce its temperature slightly. This emphasizes the warm-cool contrast in the room.

Add a subtle vignette.

Real photographic lenses dim at the edges. A vignette of 8 to 15 strength simulates that dimming and pulls the viewer's eye toward the architectural focus. Too strong looks Instagram-filter; too weak doesn't register.

Grain. Always grain.

Renders are mathematically clean. Photographs are not. Adding film grain at 1 to 3% intensity is the single fastest way to make a render look like it was taken with a camera. If you only do one of these moves, do this one.

A specific example

For the dark living room render shown at the top of this post, my actual node tree was:

1. Exposure +0.15 EV (the original render came out slightly underlit from the side window).
2. White balance: temp +200, tint -8 (push the room slightly warm, neutralize the green tint that the brass fixtures introduced).
3. Lift +0.04, gamma +0.02, gain -0.02 (lift the deep shadows under the hanging seat, slightly compress highlights on the brass).
4. Color wheels: midtones nudged warm (+orange), highlights nudged cool (+blue).
5. Curve: gentle S-curve, low end pulled up slightly.
6. HSL qualifier on the brass fixture: increased saturation 12%.
7. HSL qualifier on the window light: reduced saturation 8%, slight warmth.
8. Vignette strength 12, soft falloff.
9. Grain at 2.5%.
10. Output sharpen at 35%.

Total time on the grade: about 25 minutes. The render itself took a couple of minutes on an RTX 5090; the grade is most of the post-pipeline now that GPUs have collapsed render time. It is what turns a render into an image.

What to avoid

• Heavy LUTs designed for video footage. Most cinematic LUTs assume actual photographic source. Applied to renders, they amplify the cleanness and make the image look more rendered, not less.
• Pushing saturation to compensate for flat lighting. Fix the lighting upstream, not the grade.
• Strong sharpening. Renders are already pixel-clean. Mild output sharpen only.
• Different grades on different shots from the same project. Use the same powergrade as a starting point for every shot in a series. Variations should be intentional.

The 80/20 of color grading renders

If you only have time for three moves, do these:

1. Add film grain.
2. Slight S-curve for contrast.
3. Vignette.

These three moves on top of an unedited render will get you 70% of the way to a finished look. Everything else above is the last 30%.

For the upstream pipeline that produces the render itself, see interior render workflow in Blender. For why post matters as much as the render, see the practitioner's guide to archviz.