reference-to-3d
Skill2 starsChanged 3 days agoMITView on GitHub
Reconstruct Blender models from supplied reference sheets, branding templates, texture atlases, orthographic front/side/back/top views, or mascot/logo art where visual fidelity to the source is more important than a plausible generated object. Use when the user says the model must match a template, wireframe, texture pack, character sheet, mascot sheet, or brand asset exactly; also use after feedback like "does not look like the reference", "fit the texture 1:1", "wrong number of visible parts", or "compare against the template". Requires Blender MCP plus local Python with Pillow/OpenCV/numpy; pairs with blender-uv-texturing, wireframe-to-3d, blender-modeling, blender-materials, and blender-export.
1 other project uses a version of this, 1 of them edited.
Adopted by
Most recently taken up by
- keysforthewin/thisofficeedited, first seen Jul 24, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 1
- keysforthewin/thisoffice★ 099% same
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository33 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .github/copilot-instructions.mdCopilot instructions
- .cursor/rules/3d-fabric.mdcCursor rule
- 3d-fabricSkill
- animation-quality-gateSkill
- atlas-uv-fittingSkill
- blender-animationSkill
- blender-camerasSkill
- blender-exportSkill
- blender-lightingSkill
- blender-materialsSkill
- blender-modelingSkill
- blender-pro-workflowSkill
- blender-renderingSkill
- blender-skill-harmonizerSkill
- blender-uv-texturingSkill
- closed-surface-uv-coverageSkill
- contour-to-meshSkill
- fit-repair-optimizerSkill
- landmark-fit-repairSkill
- mascot-logo-reconstructionSkill
- multiview-constraint-solverSkill
- multiview-fit-loopSkill
- orbital-hud-motionSkill
- orthographic-registrationSkill
- quality-refinement-autoloopSkill
- reference-analysis-validatorSkill
- reference-look-calibrationSkill
- source-part-segmentationSkill
- text-to-blenderSkill
- texture-driven-mesh-fittingSkill
- texture-state-animationSkill
- wireframe-to-3dSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
14 rules. About 0 are common, 7 are distinctive, and 7 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
7
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 10
Use when: User supplies reference images/templates/texture atlases/orthographic views and wants a model matching them, or repeated visual …
- Distinctive: line 12
Before the first Blender call, read ../text-to-blender/references/mcp-backends.md.
- Distinctive: line 19
When the user requires true 1:1 matching, chain these skills instead of attempting another plausible modeling pass:
- Distinctive: line 27
Do not export a final asset until the validation JSON and overlay render pass the manifest thresholds.
- Very specific: line 31
If the user says the output does not match the templates/textures/wireframes, stop regular text-to-blender generation and enter this …
- Very specific: line 42
Before Blender modeling, write a small reference_manifest.json with:
- Distinctive: line 60
If the expected part count is unclear, derive it from the reference sheet with the analyzer and show the user the uncertainty instead of …
- Very specific: line 77
Use scripts/template_analyzer.py for first-pass component extraction:
- Very specific: line 87
Create a front reference plane or Image Empty in the same orthographic camera used for validation.
- Very specific: line 96
Add only shallow Y-depth/crown/extrusion after the front silhouette is locked.
- Very specific: line 97
Keep part names semantic: GEO-subject_part_top, GEO-subject_face_shell, etc.
- Distinctive: line 99
Do not radial-duplicate a mascot unless the reference is truly radial.
- Distinctive: line 108
If views disagree, keep the front brand read and document the conflict.
- Distinctive: line 188
Ask for a correction only after generating evidence, e.g.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
MCP backend
Use when: User supplies reference images/templates/texture atlases/orthographic views and wants a model matching them, or repeated visual mismatch feedback requires a reference-locked corrective loop.
Before the first Blender call, read ../text-to-blender/references/mcp-backends.md. Prefer official MCP for shared capabilities; preserve the selected backend for the whole task. Tool names in recipes are client/legacy examples: map them through that reference and its agent-host guide.
Reference-to-3D Reconstruction
This skill is for source-locked modeling: the reference asset is the contract. Do not generate a plausible object from memory and then decorate it. Extract measurements, part counts, silhouettes, and UV regions from the supplied files first, then build the Blender model to satisfy those measurements.
1:1 reconstruction upgrade
When the user requires true 1:1 matching, chain these skills instead of attempting another plausible modeling pass:
reference-analysis-validatorfor manifest, masks, part counts, overlay gates.orthographic-registrationfor front/side/back/top coordinate agreement.contour-to-meshfor silhouette-derived structural meshes.atlas-uv-fittingfor per-part texture/UV fitting.mascot-logo-reconstructionas the full orchestrator for brand mascot/logo work.
Do not export a final asset until the validation JSON and overlay render pass the manifest thresholds.
Triggered failure mode
If the user says the output does not match the templates/textures/wireframes, stop regular text-to-blender generation and enter this workflow. Repeated plausible-but-wrong renders usually mean the missing step is reference analysis, not another modeling iteration.
Source-of-truth hierarchy
- Front template / front wireframe: primary silhouette, visible part count, face/feature positions, brand read.
- Texture atlas: exact part shapes, color/material boundaries, decals, lightmap/aura intent.
- Side/back/top views: depth, stacking order, thickness, backside forms. They must not change the locked front projection.
- User feedback: hard constraints; promote it into validation gates.
Mandatory preflight checklist
Before Blender modeling, write a small reference_manifest.json with:
{
"source_files": [],
"primary_view": "front",
"expected_primary_parts": {"count": null, "labels": []},
"structural_parts": [],
"decorative_parts": [],
"texture_regions": [],
"validation_thresholds": {
"front_mask_iou_min": 0.90,
"bbox_center_tolerance_px": 12,
"part_count_exact": true
}
}If the expected part count is unclear, derive it from the reference sheet with the analyzer and show the user the uncertainty instead of guessing.
Pipeline
1. Inventory and classify sources
Classify each provided asset:
front_template: hero brand image or front wireframe.side_view,back_view,top_view: orthographic depth references.texture_atlas: basecolor / albedo atlas containing pieces.decal: transparent face/expression or detail layer.emissive,roughness,bump,normal,lightmap: material maps.aura/background: optional visual context, not structural body geometry.
2. Analyze images locally
Use scripts/template_analyzer.py for first-pass component extraction:
python3 ${CLAUDE_SKILL_DIR}/scripts/template_analyzer.py --image /path/to/front_reference.png --out /tmp/front_analysis.json --mode edges # use bright_on_dark/dark_on_bright/alpha when cleanerThe analyzer returns components with area, perimeter-derived score, bbox, centroid, contour, and approximate class hints. Use --mode edges for faint grey wireframes and Otsu modes for clean masks/atlases. Treat this as measurement input, not final authority.
3. Lock the front reference in Blender
Create a front reference plane or Image Empty in the same orthographic camera used for validation. Prefix it REF_ and exclude it from export. Official Blender docs describe Image Empties as reference images/blueprints and support front/back depth, opacity, orthographic-only display, and axis-aligned display.
4. Build geometry from source contours
For each structural part:
- Convert the 2D contour to front-view X/Z coordinates.
- Generate a quad strip/grid inside that contour.
- Assign Project-from-View-style UVs based on front X/Z bounds or exact atlas region.
- Add only shallow Y-depth/crown/extrusion after the front silhouette is locked.
- Keep part names semantic:
GEO-subject_part_top,GEO-subject_face_shell, etc.
Do not radial-duplicate a mascot unless the reference is truly radial. Logos often have occlusion, intentional asymmetry, and a fixed visible part count.
5. Multi-view registration
- Front X/Z is locked first.
- Side view controls Y/Z depth envelope only.
- Top view controls X/Y spread only.
- Back view controls hidden/backside geometry and validation, not front silhouette.
If views disagree, keep the front brand read and document the conflict.
6. UV and texture mapping
Chain-load blender-uv-texturing when any texture pack is provided. For atlas-driven parts:
- crop or map per atlas region; do not project the whole atlas onto every part;
- front-facing hero surfaces use front-projected UVs;
- side/back surfaces use side/back UV islands, procedural fill, or baked colors;
- alpha decals need transparent material settings and must not introduce black planes;
- lightmaps are optional realtime helpers and should not be connected as emission unless specified.
7. Validation gates before export
Always produce:
front_preview.pngfront_overlay_reference.pngfront_mask_validation.jsonside_preview.pngback_preview.pngtop_preview.png
Use scripts/silhouette_validator.py:
python3 ${CLAUDE_SKILL_DIR}/scripts/silhouette_validator.py --reference /tmp/reference_mask.png --render /tmp/render_mask.png --out /tmp/validation.jsonRefuse final export if:
- primary part count differs from manifest;
- front-mask IoU is below threshold;
- face/feature centroid is outside tolerance;
- optional aura/decorations are included in the base GLB by mistake.
Blender code patterns
Reference plane in front-view X/Z
import bpy
def create_reference_plane(name, image_path, width, height, y=0.2, alpha=0.35):
img = bpy.data.images.load(image_path, check_existing=True)
mat = bpy.data.materials.new('MAT-' + name)
mat.use_nodes = True
nodes = mat.node_tree.nodes
bsdf = nodes.get('Principled BSDF')
tex = nodes.new('ShaderNodeTexImage')
tex.image = img
mat.node_tree.links.new(tex.outputs['Color'], bsdf.inputs['Base Color'])
mat.node_tree.links.new(tex.outputs['Alpha'], bsdf.inputs['Alpha'])
bsdf.inputs['Alpha'].default_value = alpha
mat.blend_method = 'BLEND'
verts = [(-width/2,y,-height/2),(width/2,y,-height/2),(width/2,y,height/2),(-width/2,y,height/2)]
mesh = bpy.data.meshes.new(name + 'Mesh')
mesh.from_pydata(verts, [], [(0,1,2,3)])
mesh.update()
uv = mesh.uv_layers.new(name='UV')
for li, uvco in zip(mesh.polygons[0].loop_indices, [(0,0),(1,0),(1,1),(0,1)]):
uv.data[li].uv = uvco
obj = bpy.data.objects.new('REF-' + name, mesh)
bpy.context.collection.objects.link(obj)
obj.data.materials.append(mat)
obj.hide_render = True
return objValidation print contract
print('VALIDATE expected_primary_parts=', expected, 'actual=', actual)
print('VALIDATE front_locked=True reference_overlay_rendered=True')
if actual != expected:
raise RuntimeError('Part-count mismatch; refusing export')When to stop and ask for human input
Ask for a correction only after generating evidence, e.g. an overlay image or JSON showing ambiguity. Do not ask vaguely. Example: “The analyzer finds N prominent structural components plus M faint background/decorative shapes. Should the GLB include only the structural components?”
Sources distilled
- Blender Manual: Image Empties for reference images/blueprints; front/back display, opacity, orthographic and axis-aligned controls.
- Blender Manual: Trace Image to Grease Pencil works best from manually prepared black/white images and controlled resolution.
- Blender Manual: Shrinkwrap moves vertices to a target surface and is useful after silhouette lock for conforming secondary details.
- Blender Manual: UV/Image Texture and Project-from-View-style workflows require UV maps and material nodes for renders/exports.
- OpenCV docs: contour moments, area, perimeter, bounding boxes, template matching.
- scikit-image docs: SSIM for image similarity when pixel-MSE is not perceptually meaningful.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

