Stylized Lighting Budgets on Steam Deck - Why Your Unity URP Scene Looks Crunchy on Handheld and How to Fix It in One Pass (2026)

Kitsune Elements artwork - hero image for the 2026 stylized URP lighting budget guide on Steam Deck

Why this matters now

If you have ever felt the pit-of-stomach moment of opening your stylized Unity URP scene on a Steam Deck for the first time and watching the lush, painterly look you spent three months tuning suddenly look blown-out, banded, and noisy on the handheld - you are not alone. The autumn 2026 Deck Verified refresh specifically flagged stylized URP scenes for blown-out highlights and broken shadowmap precision more than any previous review wave. The reviewer image now runs an automated check that screens for HDR clipping in the top 8% of bright pixels and shadow acne above a per-frame threshold.

The good news: this is almost always fixable in one focused evening if you know which four lighting budgets to tighten and in what order. The bad news: most tutorials assume you are running on a desktop GPU with headroom to brute-force the look, and the Steam Deck simply does not have that headroom.

This post is the beginner-first companion to the Steam Deck Verified Autumn 2026 Refresh blog. It walks the four budgets that decide whether your scene reads as polished or crunchy on the Deck:

Bloom budget - intensity, threshold, soft knee, and texture format.
Exposure and tonemap budget - HDR midpoint, tonemapping curve, white point.
Shadow budget - cascade count, shadowmap resolution, normal bias, distance.
Global Illumination and probe budget - probe density, light layer mask, indirect intensity.

You will measure each before changing, change one thing, measure again. By the end of an evening, your scene will read on the Deck like you intended it to read on your desktop monitor.

Who this guide is for

Beginner-to-intermediate Unity URP users shipping a stylized 2D or 3D game.
Teams whose scene looks great on a 27-inch monitor and noticeably worse on the Deck's 7.4-inch screen.
Teams targeting Deck Verified under the autumn 2026 refresh rules.
Teams whose Deck QA reviewer wrote "blown-out" or "banded" or "shadow acne" in feedback notes.

Not for: HDRP teams (the Bloom/exposure portions translate but the URP-specific section names do not), or teams running custom render features that bypass the URP volume stack entirely. For HDRP, swap "Volume" for "Volume Profile" and the math holds; the specific URP property names will differ.

What "crunchy" means specifically - five visual symptoms

Before fixing anything, name the symptom. Crunchy is a feeling; the underlying problems are five specific failure modes:

Blown-out highlights. Bright pixels clip to pure white. Detail at the top of the exposure range is destroyed. Most common in scenes with strong directional sunlight or bright fire/lava VFX.
Banding in gradients. Smooth color transitions (sunset sky, gradient floors) show visible stair-stepping. Indicates 8-bit texture format or wrong color space.
Shadow acne. Dark stipple patterns appear on lit surfaces near shadow boundaries. Indicates normal bias too low for the shadowmap resolution and distance.
Bloom haloing. Bright objects develop fuzzy auras that bleed across silhouettes, especially against dark backgrounds. Indicates Bloom intensity too high, soft knee too aggressive, or Bloom texture format too low.
Indirect light flicker. Light Probes pop visibly as the camera moves between probe zones. Indicates probe density too low or interpolation disabled.

Most stylized scenes hit 2-3 of these simultaneously. Diagnosing each separately is the discipline that makes the one-pass fix possible.

The Steam Deck GPU profile - what makes it different

The Steam Deck is not a weaker desktop GPU; it is a different architecture optimization profile. Three differences matter for stylized lighting:

Bandwidth-constrained, not shader-constrained. The Deck has plenty of compute for stylized shaders, but moving full-screen HDR buffers around at 60 Hz pressures the memory bus. Bloom and tonemap passes feel cheap on desktop and expensive on Deck.
Native 1280×800 resolution with frequent dynamic resolution scaling. When the Deck's resolution drops to 1024×640 under thermal pressure, your screen-space effects rescale. Bloom radius scales with resolution; small misconfigurations become obvious at lower res.
OLED panel (newer Decks) has wider gamut and higher contrast than LCD. A scene tuned on the LCD Deck can look harsh on the OLED Deck even with identical settings. Tune to the LCD baseline first; the OLED behaves better.

The implication: stylized lighting on the Deck rewards lower intensity values, tighter clamps, and explicit budgets rather than the "crank it up and let the GPU figure it out" pattern that works on desktop.

Budget 1 - Bloom

Bloom is the single biggest contributor to "crunchy" on the Deck. Get this right and 60% of the work is done.

The four Bloom properties that matter

Open your URP Volume Profile (the one applied to your global Camera volume or your scene's main camera). Find the Bloom override.

Intensity. How bright the bloom contribution is. Default is 1.0 on most Unity templates; you want 0.3 to 0.7 for stylized scenes targeting the Deck. Higher values look great on a 27-inch monitor and blown-out on a 7.4-inch one because the per-pixel intensity is the same but the perceived brightness is concentrated.
Threshold. The HDR brightness above which a pixel contributes to Bloom. Default is 0.9; for stylized scenes raise this to 1.1 to 1.5 so only the genuinely bright pixels bloom. Lower thresholds make every bright pastel surface bloom and turn the scene into a soft mush.
Soft Knee. How smoothly the threshold transitions. Default is 0.5; for stylized scenes use 0.2 to 0.3 so the transition is sharper and the Bloom feels deliberate rather than a haze.
Texture Format (under URP Asset → Quality → HDR or via the Renderer asset). Set to R11G11B10 Float for the Deck. Lower-precision formats save memory but cause banding in the Bloom buffer that is very visible on the Deck panel.

Verification step for Bloom

Take a screenshot in your most Bloom-dependent scene on the Deck (or via Deck remote play). Open the screenshot in any image editor. Look for:

Bright pixels that retain detail (not pure white).
Halos around bright objects that have a clear inner edge (not a fuzzy mush).
No visible stair-stepping in the Bloom transition zones.

If any of those fail, revisit the four properties.

Budget 2 - Exposure and tonemap

The second-biggest contributor. Exposure and tonemap decide how HDR light gets mapped to display values; if this is wrong, everything else compounds.

The three exposure/tonemap properties that matter

In the same URP Volume Profile, find or add the Tonemapping and Color Adjustments overrides.

Tonemapping Mode. Use Neutral for stylized scenes targeting the Deck. ACES crushes shadows aggressively, which looks great on cinematic desktop scenes and crunchy on stylized handheld scenes. None removes tonemapping entirely and almost always clips highlights.
Post Exposure (under Color Adjustments). Default is 0; for stylized scenes targeting the Deck, set to -0.2 to 0. A slight negative exposure protects highlights at the cost of darkening shadows; the Deck's panel handles the darker shadows better than the brighter highlights.
HDR Output (under URP Asset → Quality → HDR). Enable HDR. Use the R11G11B10 Float intermediate format for the color buffer (same as Bloom). If HDR is disabled, you get LDR clamping at the lighting stage and lose all the headroom Bloom and Tonemap rely on.

Verification step for exposure

In your brightest scene on the Deck, sample the brightest pixel using Unity's Frame Debugger or just take a screenshot and read the RGB value. The brightest pixel should be 245-250 (out of 255), not 255. If it is 255, your highlights are clipping.

In your darkest scene, the deepest shadow pixel should be 5-15, not 0. Pure black on the Deck OLED panel reveals dust on the screen and looks dead.

Budget 3 - Shadows

The third budget. Shadow misconfigurations cause the shadow acne symptom and contribute to crunch.

The four shadow properties that matter

In your URP Asset (the one your Quality setting points at):

Main Light Shadow Resolution. Default is 2048 in many templates. For the Deck, 1024 is the sweet spot for stylized scenes. Higher resolution causes more shadow acne when normal bias is wrong; lower resolution is softer and forgives more bias misconfigurations.
Cascade Count. Default is 4. For stylized scenes on the Deck, 2 cascades is almost always enough. Each cascade is another full shadowmap render pass at framerate cost.
Max Shadow Distance. Default is 50 meters in many templates. For a stylized scene with limited camera range, 15-25 meters is plenty. Shorter distance means each shadowmap pixel covers less world area, so shadows are sharper and free of acne.
Normal Bias and Depth Bias (on the Main Light component). These are the magic numbers. For Unity URP 17.0.x, default Normal Bias is 1; for stylized scenes on the Deck, 0.4 to 0.8 with Depth Bias at 0.5 to 1.0 is the sweet spot. Lower Normal Bias gives crisper shadow contact; higher Normal Bias eliminates acne at the cost of "peter panning" (the shadow detaching from the object).

Verification step for shadows

Stand your camera near a wall with a small object casting a shadow at a glancing angle. Look at the boundary where the shadow meets the lit surface:

Shadow boundary should be a clean curve with no stipple pattern.
Shadow contact at the object's feet should be tight, not floating.
Camera should pan slowly without shadow flicker or popping cascade boundaries.

If you see acne, raise Normal Bias by 0.1 increments until clean. If you see peter panning, lower Normal Bias by 0.1.

Budget 4 - Global Illumination and Light Probes

The smallest budget for typical stylized scenes, but the one most beginners miscalibrate.

The three GI/probe properties that matter

Probe Density. Open the Light Probe Group in your scene. For stylized scenes targeting the Deck, place probes at 1.5-3 meter spacing in playable areas. Tighter density looks better at the cost of bake time and memory; wider density causes probe-popping flicker. Place extra probes at room transitions where indirect light changes sharply.
Indirect Intensity (under Lighting Settings). Default is 1.0; for stylized scenes the 0.5 to 0.8 range is safer. The Deck's panel exaggerates indirect light contributions; lower values restore the directional clarity stylized scenes depend on.
Probe Interpolation. Enable Use Light Probes on every Renderer that should receive indirect light. New beginners often skip this and wonder why their character glows with the wrong color in some rooms.

For Adaptive Probe Volumes (APV) in Unity 6 URP, swap the probe-spacing rule for the APV sampling-density setting and target medium quality. Higher quality looks the same on the Deck and costs more bake time and runtime memory.

Verification step for GI/probes

Walk your character through every room transition in the scene. Watch the character's skin or main material color. It should shift smoothly between rooms, not pop suddenly. If you see popping, the probes are too sparse near the transition or interpolation is off.

The one-pass remediation walkthrough (45-90 minutes)

This is the focused evening fix. Do not bundle changes; do one budget at a time and verify before moving to the next.

Step 1 - Take "before" screenshots. On the Deck (or Deck remote play), capture three screenshots: your most Bloom-dependent scene, your most shadow-dependent scene, and a room transition that crosses light probe zones. Save to lighting-audit-before/.

Step 2 - Fix Bloom (10-15 minutes).

Set Intensity to 0.5 as a starting point.
Set Threshold to 1.2.
Set Soft Knee to 0.25.
Confirm R11G11B10 Float in URP Asset.
Take a new Bloom screenshot and compare.

Step 3 - Fix exposure and tonemap (5-10 minutes).

Tonemapping Mode → Neutral.
Post Exposure → -0.1.
Confirm HDR enabled.
Take a new screenshot with the brightest pixel in frame and verify it lands around 245-250.

Step 4 - Fix shadows (15-25 minutes).

Shadow Resolution → 1024.
Cascade Count → 2.
Max Shadow Distance → 20.
Normal Bias → 0.6.
Depth Bias → 0.75.
Walk through your shadow-dependent scene and verify clean shadow contact.

Step 5 - Fix GI/probes (10-15 minutes).

Audit probe spacing in playable areas. Add probes at room transitions if needed.
Indirect Intensity → 0.7.
Verify Use Light Probes on key Renderers.
Re-bake lighting (this is the slow step; budget time for it).

Step 6 - Take "after" screenshots. Same three scenes as Step 1. Save to lighting-audit-after/. Diff visually.

Step 7 - Run the Deck smoke test. Play 5 minutes on the actual Deck (or Deck remote play). Watch for any of the five symptoms returning. If one does, identify which budget regressed and revisit that budget only.

Verification - what "fixed" looks like

After the one-pass fix, your Deck screenshots should show:

Bright pixels with detail at the top of the exposure range (245-250 max, not 255).
Smooth gradients without banding.
Shadow boundaries as clean curves with tight contact.
Bloom halos with clear inner edges, deliberate rather than mushy.
Smooth probe transitions during walks between rooms.
No visible shadow popping during camera pans.

If three out of five hold after one pass, you have shipped the fix; the remaining two usually need a second focused evening or are scene-specific edge cases that are not crunch but stylistic choices.

Common beginner mistakes

Tuning on your laptop and assuming it transfers to the Deck. It does not. Tune on the Deck (LCD baseline) or via remote play.
Cranking Bloom Intensity to 1.5 because "stylized = strong vibe." On the Deck, that vibe reads as crunch. Stylized strength comes from threshold and color, not brute intensity.
Using ACES tonemapping because tutorials use it. ACES is cinematic; it crushes shadows. Stylized scenes need Neutral on the Deck.
Disabling HDR to "save performance." This destroys the headroom Bloom and Tonemap need. Keep HDR enabled.
Maxing shadowmap resolution to fix acne. Higher resolution shifts the acne to different positions; it does not fix the underlying Normal Bias issue.
Setting Cascade Count to 4 because "more is more." The Deck pays the framerate cost; you do not see the visual benefit at the Deck's render distances.
Forgetting to enable Use Light Probes on receivers. The probes bake correctly but no Renderer reads them. Result: characters glow with wrong color in rooms.
Re-baking GI without saving the before state. If the new bake looks worse, you cannot easily roll back. Keep the previous bake's *.lighting files until verified.

Pro tips for keeping it clean

Pin the Volume Profile values in a LightingBudget.cs ScriptableObject so future scene authors do not silently raise Bloom Intensity to 1.0 again.
Set up a Deck remote-play workstation early. SteamLink + Deck running your game is the fastest tuning loop short of physical-hardware QA.
Save before/after screenshot pairs in version control. A 6-month-later "we used to look better" review needs the receipts.
Profile on Deck with the GPU Usage stat overlay to confirm Bloom and shadow passes fit within your frame budget. If the post-processing pass spikes above 4ms on the Deck, revisit Bloom intensity.
Test on both Deck LCD and Deck OLED if you have access. Tune to the LCD baseline; the OLED behaves better.
One person owns the lighting budget. Multiple contributors editing the same Volume Profile is the most common source of regression.
Audit per quarterly Deck thermal load test. Frame rates degrade under sustained heat; the lighting cost matters more in hour 2 of play than hour 1.

For the broader Deck performance picture, pair this with Steam Deck TDP Spikes vs Frame Pacing. For 2D-specific lighting context, see Lighting 2D Action Game Silhouettes - Rim, Ambient, Shader Basics and Top-Down Dungeon Lighting 2026 - 12 Contrast Rules.

URP versus HDRP - quick note

This post is URP-centric because URP is the default stylized indie pipeline on Unity 6 / 6.6 LTS. If you are on HDRP:

Bloom lives in HDRP's Volume override, same property names, similar tuning ranges but Threshold scale differs.
Exposure and tonemap are HDRP-native and richer; the Auto Exposure mode is your friend on the Deck.
Shadows in HDRP have Contact Shadows and screen-space variants; for the Deck, disable Contact Shadows for stylized scenes (the cost is not worth the visual gain).
Light Probes are replaced by APV in modern HDRP; the medium-quality rule still holds.

HDRP on the Deck is generally heavier than URP and the Deck Verified cert lane is friendlier to URP for new submissions. If you have not committed to HDRP for non-Deck reasons, URP is the more Deck-friendly choice.

When stylization should win versus when cleanliness should win

A pragmatic rule of thumb:

If the stylistic choice (high contrast, heavy bloom, deep shadows) is intentional and the team can articulate why, keep it - but tune the values until it reads on the Deck the way you intended on desktop.
If the stylistic choice emerged from default settings nobody changed, the budget rules above almost always improve the look.
Stylization should not require per-platform asset swaps. If the look needs different intensities, exposures, or shadow distances on Deck vs desktop, set per-quality-level overrides in URP Asset; do not duplicate assets.

The point of the budget framework is to make stylistic choices deliberate, not to flatten every scene to a clean Neutral look.

Key takeaways

The Deck is bandwidth-constrained, not shader-constrained. Lower intensity values and tighter clamps win on handheld where they don't matter on desktop.
Bloom is the biggest crunchy contributor. Intensity 0.3-0.7, Threshold 1.1-1.5, Soft Knee 0.2-0.3, R11G11B10 Float format.
Tonemapping = Neutral, not ACES for stylized scenes on Deck. Post Exposure -0.2 to 0 protects highlights.
Shadow resolution = 1024, cascades = 2, distance = 20m, Normal Bias = 0.6 is the stylized-on-Deck sweet spot.
GI/probe spacing at 1.5-3 meters with Indirect Intensity 0.5-0.8. Extra probes at room transitions.
Tune on the LCD Deck baseline. OLED behaves better.
One-pass remediation in 45-90 minutes: before-screenshots → Bloom → exposure/tonemap → shadows → GI/probes → after-screenshots → Deck smoke test.
Save before/after screenshot pairs in version control. Lighting regressions creep back; receipts are the defense.
One person owns the lighting budget. Multiple authors editing the same Volume Profile is the primary regression source.
Stylization should be deliberate, not emergent. If a stylistic choice survives the budget framework, it is intentional; if not, it was a default nobody questioned.

FAQ

1. We are on URP 17.0.x. Does anything specific change? URP 17.0.x on Unity 6.6 LTS tightened a couple of Bloom defaults and changed the Volume override discovery path slightly. The properties named above all exist; some have moved one tab in the Inspector. If a property name does not match, the Unity URP Manual will resolve the renames in seconds.

2. We use a stylized custom Lit shader (not Lit URP). Does this still apply? Mostly yes. Bloom, exposure/tonemap, and probe interpolation are render-pipeline-level; they apply regardless of which surface shader your Renderers use. Shadow tuning is sometimes shader-specific (custom shaders can override Normal Bias or skip cascaded shadow sampling); audit your shader against the documented stylized Lit shader to confirm shadow sampling parity.

3. Our scene looks fine on Deck but the reviewer flagged "banding." What gives? Almost always one of: (a) HDR disabled in URP Asset, (b) Bloom buffer at lower precision than R11G11B10 Float, or (c) post-processing target format set to LDR before tonemap. Audit those three first.

4. Should we use Dynamic Resolution on the Deck? Yes, but tune for the minimum resolution the dynamic scaler is allowed to drop to, not the native resolution. If the scaler is allowed to drop to 1024×640, Bloom radius and shadow boundaries will look noticeably different there. Test the lowest-quality scaler setting before submission.

5. Can we automate any of this? The screenshot capture and diff steps can be CI-automated using Unity's Graphics Test Framework. The actual Volume Profile and URP Asset values cannot be auto-tuned; they need human judgment. Automate the regression detection (did Bloom Intensity drift back above 0.7?), not the initial tuning.