Lineart + Depth — Shape First, Color Later (CloudDock)

Why Lineart + Depth works

When you prompt from scratch, SD has to decide everything at once: shape, composition, lighting, and style. That’s why images drift — and why consistency can be painful.

Lineart + Depth is a more controlled workflow:

Shape first, color later.
Use Lineart to lock edges and silhouette, use Depth to lock perspective and spacing, then let the prompt “paint” style and color on top.

Best use cases

Redrawing an existing sketch into anime style
Keeping composition stable across iterations
Preventing background drift while you refine character design
Turning rough concept art into a cleaner render

The pipeline (two controls, one clean result)

Lineart controls edges and shape (the “drawing”).
Depth controls 3D layout and distance (the “stage”).
Your prompt controls style, color, mood (the “painting”).

Picture showing shape-first color-later pipeline using lineart and depth control

Structure first. Style second. Cleaner results, less drift.

What you need (inputs)

A source image (sketch, draft, or even a messy render)
ControlNet enabled in A1111
Two slots: Lineart + Depth

Tip: This pipeline works best when the source image has clear separation: readable lines + readable foreground/background. If everything is muddy, fix contrast first.

Preset: Lineart + Depth (recommended)

ControlNet #1 — Lineart (shape lock)

Weight: 0.70–0.95
Start / End: 0.00 → 0.85
Goal: keep silhouette and edges stable

ControlNet #2 — Depth (space lock)

Weight: 0.45–0.70
Start / End: 0.10 → 0.90
Goal: keep perspective and spacing consistent

Why depth is softer than lineart: Depth should guide layout, not freeze the entire render into a plastic 3D look. Let lineart do the hard “shape lock,” let depth do gentle “stage direction.”

A1111 ControlNet panel showing lineart and depth configured in two slots with different weights

Lineart strong, depth medium: stable composition without stiffness.

Prompting: how to “paint” on top

With Lineart + Depth handling structure, your prompt should focus on:

style (anime, clean lineart, cel shading, soft painterly, etc.)
lighting (sunset rim light, studio softbox, neon city)
color palette (pastel, high contrast, monochrome)
material / texture (silk, denim, glossy armor)

Example prompt:

anime illustration, clean lineart, cel shading, soft gradient background,
cinematic lighting, warm color palette, detailed eyes, sharp focus

Negative starter:

blurry, lowres, bad anatomy, extra limbs, deformed hands,
messy lines, oversharp, noisy background, worst quality

Common mistakes (and quick fixes)

“It looks stiff / traced.”

Lower Lineart weight to 0.65–0.80
End Lineart earlier (end 0.75–0.85)
Let the last steps add style detail naturally

“Depth makes it look 3D/plastic.”

Lower Depth weight to 0.40–0.55
Start Depth later (start 0.15–0.25)
Make depth influence less dominant than lineart

“Background still drifts.”

Increase Depth weight slightly (+0.05)
Use a cleaner source image (better contrast / clearer background)
Reduce denoising if you’re using img2img

VRAM notes (A-group)

Two ControlNets cost VRAM. Keep the workflow stable by staying modest:

Batch size 1
Moderate base resolutions (don’t start huge)
If you use hires, use the 1.3× ladder and low denoise

OOM survival rule: reduce base resolution first. Cutting a few steps won’t save you if the canvas is too large.

What’s next?

Pose · Hand · Face presets: hands strong, faces medium, body soft.
Go to “Pose · Hand · Face →”
Two-stage hires ladder: 1.3× first, low-noise repair second.
Go to “Two-stage Hires →”
img2img Basics: denoise logic + inpaint repair.
Go to “img2img Basics →”

Lock the shape. Paint the vibe.

← Back to SD Home