Text To Blender

Text-to-Blender Orchestrator

Turn plain-English requests into Blender work. You are the conductor: read the request, decide which sub-skills to chain-load, sequence them in the right order, and execute via the Blender MCP.

Multi-skill harmonization

For complex tasks that trigger multiple Blender skills, especially reference/template/brand work, load blender-skill-harmonizer before choosing the execution order. It owns precedence, handoff artifacts, and conflict policy.

If the user rejects an output as sub-par, says the same issue is recurring, or asks to improve the skill stack before retrying, load quality-refinement-autoloop before further product work. It turns evidence into a sanitized reusable lesson, patches generic skills/docs/versioning when explicitly requested, and only then resumes the Blender repair loop.

How this skill works

The user speaks in tasks ("render a hero shot of a sword on a stone"); you:

1. Identify intent → which capabilities are needed (modeling? materials? lighting? rendering? export?).
2. Check prerequisites → MCP server reachable, scene state.
3. Reset world to a known-good baseline (see World reset below) — prevents leftover broken HDRI / Environment Texture nodes from previous runs corrupting the render.
4. Look up real-world dimensions for the subject (references/common-object-dimensions.md) BEFORE generating Python. A "sword" has specific proportions; a "chair" has specific proportions; do not guess.
5. Route to sub-skills → load the relevant ones via Read and follow their instructions.
6. Sequence the work in the order pros use (see references/assembly-order.md).
7. Execute generated Python via mcp__blender__execute_blender_code.
8. Validate with mcp__blender__get_scene_info, get_object_info, AND `get_viewport_screenshot` — see Visual validation checkpoint below. Numerical validation alone is not enough: the API can report success while geometry is grossly wrong.
9. Iterate — if the visual check shows an obvious problem (subject barely visible, wrong proportions, wrong orientation), fix and re-render BEFORE reporting success.
10. Report to the user with concrete numbers AND a path to the proof render they can inspect.

World reset (always step 3, before any composition)

Previous tests can leave the scene's world in a broken state — particularly an Environment Texture node with image=None cascading into Background, which produces a magenta-flooded render. Reset world to a neutral baseline at the start of every scene-build:

def reset_world(scene, color=(0.04, 0.04, 0.05, 1.0), strength=0.4):
    world = scene.world
    world.use_nodes = True
    nodes = world.node_tree.nodes
    for n in list(nodes):
        nodes.remove(n)
    output = nodes.new('ShaderNodeOutputWorld')
    bg = nodes.new('ShaderNodeBackground')
    bg.inputs['Color'].default_value = color
    bg.inputs['Strength'].default_value = strength
    world.node_tree.links.new(bg.outputs['Background'], output.inputs['Surface'])

If the user wants HDRI lighting, override this AFTER it runs (load the actual .hdr file and verify env.image is not None).

Visual validation checkpoint (always between steps 8 and 10)

After rendering, always:

1. Call mcp__blender__get_viewport_screenshot (or read the rendered file with Read).
2. Visually verify against the user's request. Specifically:

• Subject is visible and recognisable (not a thin streak, not magenta-flooded, not entirely in shadow)
• Proportions match the real-world reference dimensions for that subject
• Composition is reasonable (subject in frame, not clipped at edges, not microscopic in one corner)
• Materials have visible variation (not perfectly flat plastic-looking surfaces) — for production-quality scenes, add procedural texture nodes

1. If the render is obviously wrong, do NOT report success. Iterate: identify the specific problem, fix it, re-render, re-check.

A render passing every numerical check (object count, vertex count, file size, no error messages) can still look completely wrong. The user only sees the picture. Make sure the picture matches the request before declaring done.

Set viewport to Material Preview mode (last step before reporting)

The user is often looking at Blender's viewport, not the rendered file. Blender defaults the viewport to Solid shading mode which ignores all materials and shows everything as flat grey. After scene assembly, switch the viewport to Material Preview so the user actually sees the materials they got:

import bpy

for area in bpy.context.screen.areas:
    if area.type == 'VIEW_3D':
        for space in area.spaces:
            if space.type == 'VIEW_3D':
                space.shading.type = 'MATERIAL'           # or 'RENDERED' for full quality
                space.shading.use_scene_lights = True
                space.shading.use_scene_world = True

Skip this only when the user has explicitly asked you to leave the viewport alone.

Real-world dimension lookup

Before sizing any subject from natural language, consult references/common-object-dimensions.md. It lists realistic proportions for common requests:

The full reference list lives in references/common-object-dimensions.md. Do not guess.

Prerequisites — always check first

Before any work, verify:

1. Blender MCP is reachable. Call mcp__blender__get_scene_info. If it errors with "Could not connect to Blender": > "Blender's MCP addon isn't running. Start Blender, enable the BlenderMCP addon (port 9876 default), then re-run."

Stop and ask the user to fix this.

1. Scene state. get_scene_info returns the current objects. Decide:

• Empty scene? → Start fresh; build from primitives.
• Existing objects? → Operate on them; do NOT delete unless asked.
• Default cube only? → Probably safe to delete (bpy.ops.object.delete()).

1. The user's actual request. If ambiguous, ask one question. Otherwise proceed with sensible defaults.

Intent → sub-skill routing table

For each intent the user expresses, load (via Read) the matching sub-skill's SKILL.md and follow its patterns. Multiple intents = chain multiple sub-skills.

Multi-intent example: "Model a sword with materials, light it dramatically, and export as glTF" →

1. blender-pro-workflow (sequencing strategy)