Blender Export logo

Blender Export

Community
RobLe3
blender-export

Export Blender scenes to glTF/GLB (web/AR), FBX (game engines), OBJ (universal), USD (VFX pipelines), STL (3D printing). Includes Unreal collision hulls and Multires normal-map baking for low-poly export, per-format settings, embed/unpack textures, axis conversion, polygon optimization (Decimate), and target-platform validation. Use whenever the user asks to "export this", "save as glTF / FBX / OBJ / STL / USD", "package for Unity / Unreal / Three.js / web / AR / 3D print", or any output format conversion. Make sure to use this skill even if the user does not say "export" — also covers "package this for the web", "make it work in Unity", "send to Unreal", "save for 3D printing".

Overview

PublisherRobLe3
Repositorycc-blender-skill
Skill nameblender-export
Stars
131
Forks
14
Bundled files
5
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 5 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by RobLe3 on GitHub. Read the source before you install it.

Installation

Install the Blender Export AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/RobLe3/cc-blender-skill/tree/main/plugin/skills/blender-export
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/RobLe3/cc-blender-skill.git /tmp/cc-blender-skill
mkdir -p .claude/skills
cp -r /tmp/cc-blender-skill/plugin/skills/blender-export .claude/skills/blender-export
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Blender Export in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Blender Export on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Blender Export is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

MCP backend

Use when: Any export to a non-.blend format, packaging for game engines, web, AR, or 3D printing.

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.

Blender Export

Export to the right format with the right settings. Wrong format choice = days of debugging in the target platform.

Format decision tree

Where is this going?
├── Web (Three.js, Babylon.js, model-viewer, AR Quick Look) → glTF / GLB
├── Game engine (Unity, Unreal, Godot)
│   ├── Animated/rigged → FBX (or glTF for modern engines)
│   └── Static → OBJ or FBX or glTF
├── Apple AR (USDZ) → USDZ (special, see Recipe 6)
├── 3D printing → STL (geometry only, must be watertight)
├── VFX pipeline (Maya, Houdini, Nuke) → USD
└── DCC roundtrip → FBX (industry standard)

Quick rule for unknown target: glTF / GLB. Open standard, modern, universally supported.

Recipes

Recipe 1 — glTF / GLB export (web / AR / general)

python
import bpy

bpy.ops.export_scene.gltf(
    filepath='/tmp/output.glb',
    export_format='GLB',                 # single-file binary; preferred
    export_apply=True,                   # apply modifiers before export
    export_materials='EXPORT',
    export_image_format='AUTO',          # PNG; AUTO falls back to JPEG for opaque images
    export_yup=True,                     # Y-up convention (most engines / web expect this)
    export_animations=True,              # toggle off for static models
    export_morph=True,                   # shape keys
    export_skins=True,                   # armatures + weights
    export_normals=True,
    export_tangents=False,               # skip unless target uses tangent-space normals beyond standard
)

# Verify
import os
size_mb = os.path.getsize('/tmp/output.glb') / (1024 * 1024)
print(f"export:gltf {size_mb:.2f} MB")

glTF caveats:

  • Only Principled BSDF materials export cleanly. Procedural shaders are dropped or simplified.
  • Hard cap: 15 MB; soft target: 8 MB.
  • No KTX2 / Draco compression (unless target supports those loaders).
  • PNG textures only (max 1024×1024 typical).

Recipe 2 — Decimate before export (if too large)

python
import bpy

obj = bpy.data.objects['GEO-target']
if obj.type != 'MESH':
    raise ValueError('Decimation requires a mesh')
selected = list(bpy.context.selected_objects)
active = bpy.context.view_layer.objects.active
mode = active.mode if active else 'OBJECT'
try:
    if mode != 'OBJECT':
        bpy.ops.object.mode_set(mode='OBJECT')
    bpy.ops.object.select_all(action='DESELECT')
    obj.select_set(True)
    bpy.context.view_layer.objects.active = obj
    mod = obj.modifiers.new('Decimate', type='DECIMATE')
    mod.decimate_type = 'COLLAPSE'
    mod.ratio = 0.7    # keep 70% of faces; lower = more reduction
    bpy.ops.object.modifier_apply(modifier=mod.name)
finally:
    bpy.ops.object.select_all(action='DESELECT')
    for item in selected:
        item.select_set(True)
    bpy.context.view_layer.objects.active = active
    if mode != 'OBJECT':
        bpy.ops.object.mode_set(mode=mode)

print(f"decimated:{obj.name} verts:{len(obj.data.vertices)}")

Then re-export. Iterate ratio until file size fits target.

Recipe 3 — FBX export (game engines)

python
import bpy

bpy.ops.export_scene.fbx(
    filepath='/tmp/output.fbx',
    use_selection=False,
    apply_unit_scale=True,
    apply_scale_options='FBX_SCALE_ALL',
    bake_space_transform=True,        # critical: applies rotation to mesh
    object_types={'MESH', 'ARMATURE', 'EMPTY'},
    use_mesh_modifiers=True,
    mesh_smooth_type='FACE',
    use_armature_deform_only=True,
    bake_anim=True,
    bake_anim_use_all_bones=True,
    bake_anim_use_nla_strips=True,
    bake_anim_use_all_actions=True,
    bake_anim_force_startend_keying=True,
    embed_textures=True,              # critical: embed textures into FBX
    path_mode='COPY',
    axis_forward='-Z',
    axis_up='Y',
)
print('export:fbx')

Common FBX gotchas:

  • Model rotated 90° in target → check axis_up='Y', axis_forward='-Z'
  • Model 100× too large → Apply Transform on the object before export
  • Textures missing → embed_textures=True, path_mode='COPY'
  • Animation only plays one action → use NLA + bake all actions

Recipe 4 — OBJ export (simple / universal)

python
import bpy

bpy.ops.wm.obj_export(
    filepath='/tmp/output.obj',
    export_animation=False,
    apply_modifiers=True,
    export_eval_mode='DAG_EVAL_VIEWPORT',
    export_uv=True,
    export_normals=True,
    export_materials=True,
    export_triangulated_mesh=False,
    forward_axis='NEGATIVE_Z',
    up_axis='Y',
)
print('export:obj')

OBJ has no animation, no rigging, basic material support only. Use for simple geometry exchange.

Recipe 5 — STL export (3D printing)

python
import bpy

bpy.ops.wm.stl_export(
    filepath='/tmp/output.stl',
    ascii_format=False,    # binary STL (smaller, faster)
    apply_modifiers=True,
)
print('export:stl')

Critical for STL:

  • Mesh must be watertight (no holes, no flipped normals, no internal faces).
  • Pre-export: in Edit Mode, run Mesh → Clean Up → Make Manifold.
  • STL has NO units — most slicers assume mm. Set Blender scene units to mm before modeling.
  • STL has NO color/material — single-color grey only.

Recipe 6 — USD export (VFX pipeline)

python
import bpy

bpy.ops.wm.usd_export(
    filepath='/tmp/scene.usdc',
    export_animation=True,
    export_uvmaps=True,
    export_normals=True,
    export_materials=True,
    use_instancing=True,
    export_textures=True,
    overwrite_textures=True,
)
print('export:usd')

USD variants:

  • .usd — text-based (debuggable, large)
  • .usdc — binary (compact, fast — default choice)
  • .usda — ASCII (human-readable, larger)
  • .usdz — zipped USD with all assets (Apple AR / iOS)

Recipe 7 — Pre-export checklist (run before any export)

python
import bpy

# 1. Apply transforms (rotation + scale baked into geometry)
obj = bpy.data.objects['GEO-target']
bpy.context.view_layer.objects.active = obj
bpy.ops.object.transform_apply(location=False, rotation=True, scale=True)

# 2. Recompute normals
bpy.ops.object.mode_set(mode='EDIT')
bpy.ops.mesh.select_all(action='SELECT')
bpy.ops.mesh.normals_make_consistent(inside=False)
bpy.ops.mesh.remove_doubles(threshold=0.0001)
bpy.ops.object.mode_set(mode='OBJECT')

# 3. Apply modifiers (some exporters keep them, but for portability apply before export)
# Already done via export_apply=True / use_mesh_modifiers=True flags

# 4. Verify mesh stats
mesh = obj.data
print(f"export_check:{obj.name} verts:{len(mesh.vertices)} faces:{len(mesh.polygons)}")

Recipe 8 — Verify exported file

python
import os

filepath = '/tmp/output.glb'
if not os.path.exists(filepath):
    print(f"ERROR:not_found {filepath}")
else:
    size_mb = os.path.getsize(filepath) / (1024 * 1024)
    print(f"verified:{filepath} {size_mb:.2f}MB")

Always verify after export. Use the host's shell tool: ls -la /tmp/output.glb.

Recipe 9 — Unreal collision meshes

Export a separate collider with the render mesh in the same FBX. Its name must match the render mesh exactly: SM_Chair → UCX_SM_Chair_01, _02, and so on. For existing GEO- names, derive the collider name from that actual name; do not rename the render mesh just to fit an example.

PrefixRequired shape
UCX_One closed, outward-wound convex hull
UBX_Rectangular box, not deformed geometry
USP_Sphere, not a stretched ellipsoid
UCP_Cylinder with hemispherical ends

Use several simple hulls for concave props: one hull around a table blocks the space underneath. Derive positions from geometry, not hand-entered guesses. A face-plane convexity test alone accepts open or empty meshes; it is not an export gate. Use scripts/collision_preflight.py to inspect the evaluated collider, check names/closedness/winding/volume/convexity, and export only the intended objects. It rejects rather than repairs invalid geometry and preserves selection/mode.

python
import bpy, runpy
# Substitute the installed helper path; load inside Blender, not ordinary Python.
helpers = runpy.run_path('/absolute/path/to/blender-export/scripts/collision_preflight.py')
mesh = bpy.data.objects['SM_Chair']
colliders = [bpy.data.objects['UCX_SM_Chair_01']]
result = helpers['export_collision_fbx'](mesh, colliders, '/tmp/chair_collision.fbx')
print(result)

The primitive checks cover local-axis authored boxes/spheres/capsules; use UCX_ for shapes outside those constraints. Boxes/spheres are unsuitable for a non-uniformly scaled render object. Use a new output path rather than overwriting. Only one render mesh per collision FBX is supported by this helper.

Engine-side behavior is not yet validated. Epic documents these conventions, but an Unreal import must verify hull count, openings, scale, and collision before claiming engine acceptance. Missing custom collision may instead trigger automatic collision generation; do not assume every importer configuration behaves alike.

Recipe 10 — Multires: bake to the cage, then export

A live Multires modifier may export every evaluated subdivided face or omit sculpt geometry, depending on exporter settings. For a low-poly asset, bake the detailed surface to a normal map and export level 0. A 30-face cage at level 8 is approximately 2 million faces.

Use scripts/multires_bake_export.py for a single-material target with UVs, an active Image Texture holding a Non-Color bake image, and exactly one enabled Multires modifier last in the stack. The helper sets viewport level 0 before baking, keeps it there through FBX export, saves/connects the normal map, and then restores selection, active object, mode, bake settings, engine, and subdivision levels. Use a single-user material and bake image. It does not overwrite an existing normal connection or accepted output files.

python
import bpy, runpy
helpers = runpy.run_path('/absolute/path/to/blender-export/scripts/multires_bake_export.py')
result = helpers['bake_export_cage'](
    bpy.data.objects['GEO-target'], '/tmp/target_normal.png', '/tmp/target_cage.fbx')
print(result)

A saved normal map/material connection is the intended change. On failure, generated files remain failure evidence; the helper restores temporary state and removes its new normal connection. Use new paths for a retry. To include colliders, pass an export_fbx callback that calls export_collision_fbx inside this helper's level-0 context; see references/unreal-export.md. Do not re-export after subdivision levels have been restored, or add ordinary meshes to the bake selection.

The helper detects the installed 5.x/4.x bake-property layout rather than assuming a spelling. Live validation is on Blender 5.1.1; Blender 4.x and Unreal imports remain unverified. Normal maps preserve surface shading, not silhouette changes.

Polycount targets per platform

PlatformTargetNotes
Web (glTF)≤ 30 000 trisMobile-safe
Hero web asset≤ 60 000 trisDesktop OK
Unity / Unreal hero50 000–100 000 trisHigh-end
Mobile game≤ 10 000 trisPer asset
AR USDZ≤ 50 000 trisiOS recommendation
3D printunlimitedBut export size matters

Common pitfalls

SymptomFix
FBX has no texturesSet embed_textures=True, path_mode='COPY'
Procedural material missing in glTFBake to image textures first; use only Principled BSDF
Game engine: model rotated 90°FBX: axis_up='Y', axis_forward='-Z'; glTF: export_yup=True
Game engine: model 100× too largeApply Transform; check unit scale
OBJ won't import elsewhereStick to ASCII filenames
STL won't printMake Manifold; recompute normals
GLB > 15 MBApply Decimate (Recipe 2); reduce textures to 1024×1024
Bone count exceededLimit weights to 4 per vertex; reduce bone count
Animation didn't exportglTF: export_animations=True; FBX: bake_anim=True
Unreal: missing/wrong custom collisionCheck names, convex closed hulls, and same-file selection (Recipe 9); inspect the import in Unreal
Multires bake fails with a misleading multi-resolution errorKeep Multires last and isolate the target selection (Recipe 10)
Baked normals disagree with exported cageSet viewport level 0 before baking, not just before export
FBX unexpectedly contains millions of facesViewport levels controls evaluated Multires export, not total_levels
Unreal: no smoothing-group informationUse mesh_smooth_type='FACE'; then verify inside Unreal

When to load references/overview.md

Load when:

  • Multi-format batch export needed
  • Asset Browser / library override workflow
  • USDZ for Apple AR specifics
  • Game engine roundtrip troubleshooting (Unity/Unreal-specific quirks)
  • LOD generation strategy

The reference covers: per-format pitfalls, asset browser workflow, library overrides for production, USD composition, polycount targets per platform.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Blender Export AI skill do?

Export Blender scenes to glTF/GLB (web/AR), FBX (game engines), OBJ (universal), USD (VFX pipelines), STL (3D printing). Includes Unreal collision hulls and Multires normal-map baking for low-poly export, per-format settings, embed/unpack textures, axis conversion, polygon optimization (Decimate), and target-platform validation. Use whenever the user asks to "export this", "save as glTF / FBX / OBJ / STL / USD", "package for Unity / Unreal / Three.js / web / AR / 3D print", or any output format conversion. Make sure to use this skill even if the user does not say "export" — also covers "pac...

Why use Blender Export on TypingMind?

Because you install it once and use it with any model. Blender Export is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Blender Export in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/RobLe3/cc-blender-skill/tree/main/plugin/skills/blender-export. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Blender Export?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Blender Export?

As many as you like. As long as a model supports skills, you can use Blender Export with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Blender Export AI skill free?

Yes. It is published on GitHub by RobLe3 under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇