MAX-786/claude-3d-harness / library/cc/text-to-blender
text-to-blender
VerifiedDrive Blender from natural language. Converts plain-English requests ("model a sword and render it with cinematic lighting", "make this glass look frosted", "set up three-point lighting", "export this scene as glTF for the web") into Blender Python code executed via the Blender MCP server. Acts as the orchestrator that picks and chain-loads specialised sub-skills (blender-modeling, blender-materials, blender-lighting, blender-cameras, blender-rendering, blender-animation, blender-export, wireframe-to-3d, blender-pro-workflow, blender-skill-harmonizer, quality-refinement-autoloop). Use this skill whenever the user wants Claude to do anything in Blender, including creating geometry, applying materials, lighting a scene, framing a camera, rendering, animating, or exporting. Make sure to invoke this skill even if the user does not say "Blender" — also covers requests like "create a 3D model of...", "render this...", "make a glTF from...", "set up a scene with...", or any 3D-creation task. Requires the Blender MCP addon (ahujasid/blender-mcp) running on port 9876.
references/assembly-order.md
View on GitHub# Scene Assembly Order — The Pro Sequence
The order in which you build a Blender scene determines whether you'll work efficiently or constantly redo work. This is the canonical sequence for nearly all production work.
---
## The 11-step order
```
1. Reference + plan → Note goals; gather visual references
2. Block-out → Primitives at correct scale; validate composition
3. Camera + composition lock → Pick focal length; frame the shot
4. Light pass v1 → Three-point lighting; mood without color
5. Refine geometry → Real models replacing primitives
6. Materials v1 → Flat colors; values before color
7. Light pass v2 → Add color, tune ratios
8. Detail pass → Sculpting, fine textures, polish
9. Final render → Production samples, denoising
10. Composite → Color grade, glare, vignette
11. Export / output → Target platform format
```
---
## Why this order
### Why blockout before detail?
A 30-minute primitive blockout reveals composition flaws before you've sunk hours into modeling. If the camera angle is wrong, you reshoot — easy at the blockout stage, painful after detail work.
### Why lighting before materials?
Lighting determines what you see. A material's roughness, glossiness, and texture only read correctly under appropriate lighting. Trying to match materials in flat lighting and then "lighting it later" guarantees rework.
### Why values before color?
Squint at any great image: it works in grayscale. The darkest dark + brightest bright should establish the focal point. Color is decoration on top of values. Get values right first, color tweaks later.
### Why detail last?
Detail is the most expensive work per visual impact. Don't invest in pores until you've validated the silhouette, lighting, and materials.
### Why composite last?
Compositing (color grading, glare, vignette) is non-destructive post-processing. It's the final 10% that takes a render from "good" to "polished". But it can't fix bad lighting or wrong composition.
---
## Time budgets per stage (for a hero still)
| Stage | % of total time |
|-------|-----------------|
| Reference + plan | 5–10% |
| Block-out + camera | 10–15% |
| Lighting setup | 10–20% |
| Modeling refinement | 30–40% |
| Materials + textures | 15–20% |
| Final render + composite | 5–10% |
**Anti-pattern**: 90% modeling, 10% lighting. Lighting and materials are what make a model *look* good; great geometry with bad lighting renders worse than mediocre geometry with great lighting.
---
## When to deviate from the order
The order is for **new scenes from scratch**. Common deviations:
- **Iteration on existing scene**: skip to step 5+ on the affected parts only.
- **Animation pipeline**: insert blocking at step 5 (rough timing on primitives), refine animation at step 8.
- **Asset library work**: jump to step 4 (lighting) immediately if you're just dressing a pre-built environment.
- **Tutorial recreation / reference**: follow the original's order if it makes pedagogical sense.
But: when in doubt, follow the order.
---
## Decision triggers — when to escalate fidelity
Do not move forward until current stage clears its check:
| Stage | Move on when… |
|-------|--------------|
| Block-out | Camera framing reads from thumbnail; silhouette tells the story |
| Form refinement | Forms read clearly at the rendered resolution |
| Light v1 | Subject pops from background; shadows describe form |
| Materials v1 | Grayscale values match reference; recognizable without color |
| Light v2 | Color mood matches reference; eye is led to subject |
| Detail | Within 30% time budget; value-add ratio is dropping |
| Composite | Image holds at thumbnail size AND 100% zoom |
---
## Critique protocol (every couple hours)
1. **Squint** — blur eyes / view thumbnail. Composition still readable?
2. **Greyscale** — values working independent of color?
3. **Flip horizontal** — catches blind-spots and asymmetry.
4. **Compare to reference** — side-by-side at same size.
5. **Walk away** — 30 minutes; come back fresh.
---
## Recovery patterns (when something goes wrong)
| Symptom | First check | Fix |
|---------|------------|-----|
| Render looks "flat" | Lighting variety | Add rim light, increase fill ratio |
| Render looks "too dark" | Exposure or HDRI | Bump exposure + 0.5; or HDRI strength |
| Render looks "too bright" | View transform | Switch to AgX from Standard |
| Materials look "plastic-y" | Roughness uniform | Add procedural roughness variation |
| Render looks "videogame-y" | Hard shadows + sharp geo | Soft shadows + bevel everything |
| Anatomy / proportions wrong | Reference closed | Reopen reference; compare landmarks |
| Render takes hours | Sample count | 256 + denoise; lower bounces |
| Animation jitters | Sub-frame mismatch | Bake simulations; check `frame_step` |
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |