Three ways to build a relief.
Two run locally in your browser and are free. The third — the Relief Generator — sculpts a relief from a text prompt and automatically refines it with a real cloud depth-estimation model. Pick by source material, not by habit. None of the local engines processes a real photograph — Depth2Beam detects one on upload and points you to the Relief Generator instead.
| Engine | Best for | Reads | Cost |
|---|---|---|---|
| Shape | Logos, crests, flat line art | Silhouette / distance transform | Free |
| Blend | Flat graphic artwork with real shading or texture on top | Silhouette + grayscale luminance, mixed | Free |
| Relief Generator | No source image — describe the subject | Text-to-relief diffusion model, refined with real AI Depth | 4–5 credits |
Neither local engine reads a real photograph — Depth2Beam detects one on upload and blocks Shape and Blend for that file rather than forcing it through math built for flat artwork. AI Depth and AI Depth + Detail are the cloud depth-estimation models that handle real photographic/anatomical depth instead — they're not a separate manual choice, but run automatically as part of a Relief Generator generation (its AI Depth refinement toggle), so there's nowhere to point a raw photo upload at them directly.
Shape
Distance-transform relief built for flat 2D illustrations — carves outward from the silhouette edge rather than reading grayscale tone. Ideal for logos, crests, and clean vector-style artwork.
Details →Blend
Layers a grayscale-luminance read of shading and texture on top of Shape's silhouette dome, with Detail Strength, Denoise, and a Luminance Curve to shape how brightness maps to height. For artwork that's mostly flat graphic but carries real shading or engraved-style detail — not for real photographs.
Details →Relief Generator
No source artwork required, and no upload option at all — every generation starts from a text prompt. Depth2Beam's diffusion + relief-style model carves a sculpted medallion depth map from the description, then an automatic AI Depth refinement pass (optionally with an added Surface Detail texture layer, "AI Depth + Detail") turns that into genuine spatial depth instead of a flat brightness read. The dedicated Coin Depth Maps workflow is this same engine, locked to a round medallion composition and exporting nothing but the coin-cropped height map.
4 credits, or 5 with AI Depth refinementNot sure which engine to use?
Auto Prep asks whether your artwork is Logo/Crest or Mixed and picks a sensible starting engine for you — a real photo or portrait upload is flagged and blocked instead of forced through the wrong math.
Choosing an Engine →