Blender Game Skills
A Claude Code skill that builds a game-ready 3D asset in Blender from concept art and checks each phase by measurement.
- 173 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The agent runs Blender scripts and writes asset files into the working folder.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/blender-game-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent runs Blender scripts and writes asset files into the working folder.
This is third-party code. Review the repository files before installing.
What it does
The `blender-image-to-3d` skill turns references (concept art, photos, turnarounds, sketches) into a game asset: a character, creature, building, vehicle, weapon or environment piece. The agent works through phases from brief to export and accepts each one on a render from the reference camera rather than on how the viewport looks. The output is a GLB or FBX file with a manifest, verified by re-importing it into an empty scene. Anything the references do not show is marked as inferred in the brief and the final report.
Who it is for. Indie developers and technical artists who need a rigged asset for Unity, Unreal or Godot from existing images.
Good fit when
- You have model sheets with front, side and back views and need a rigged asset from them.
- A vehicle with separate wheels and opening doors is built from one concept.
- A building photo is broken into a modular kit on a 2 m grid.
- A sword or other prop has to fit a triangle budget.
Not a fit when
- The asset is sculpted by hand without references.
- Blender is not installed on the machine.
- You only need a quick draft without proportion checks.
Example request
Here is a knight model sheet with front and side views. Build him in Blender for Unity, 1.85 m tall, rigged, with LOD0 to LOD2.Limitations
The skill runs inside Blender and drops its scripts into the Claude Code session, so a local Blender is required. A full pass through the rig, animation and baking phases takes a lot of time and tokens. Installing through `npx skills` needs Node.js 22.20 or newer and sends anonymous telemetry, which `DISABLE_TELEMETRY=1` turns off. Where the references carry no numbers the agent estimates proportions itself; those spots are marked as inferred.
How to disable. Remove the skill folder with `npx skills remove blender-image-to-3d` or delete `~/.claude/skills/blender-image-to-3d` manually. Finished asset files stay on disk.
FAQ
How do I install the skill in Claude Code?
Run `npx skills add majidmanzarpour/blender-game-skills --skill blender-image-to-3d -g -a claude-code`. Manually, copy `skills/blender-image-to-3d` into `~/.claude/skills/`.
Does the skill trigger on its own?
Yes, Claude picks it up when images are attached to a message asking for a 3D model. To force it, type `/blender-image-to-3d` followed by your request.
Which output formats does it produce?
GLB or FBX together with `asset-manifest.json` and `animation-contract.json`. After export the file is re-imported into an empty Blender scene and checked against the manifest.
What does measurement-based gating mean?
At each phase the agent compares a render with the reference by silhouette overlap (IoU) and a width profile. Detail is not added until the silhouette passes the tolerance.
Do I need to know the asset size in advance?
No. If scale is not visible in the references the agent proposes defaults (1 unit equals 1 m, Z up, GLB) and asks only for what the images do not give.
Discussion
No questions or reviews yet.
Related
SkillsA skill collection guides AI agents through Blender modeling, materials, animation and asset checks using supplied Python scripts.
Blender Skills
blender-skills
Blender skills for Cursor, Claude Code, Kiro and Codex: work goes through the BlenderMCP add-on that has to be installed in Blender
A loop skill: the agent generates a target screenshot, builds toward it, and uses a critic subagent to close the gap until they match
A skill that takes an agent from a generated picture to a game asset: sprites, video-to-frames animation and pre-release checks
Scenario Agent Skills
scenario-labs-skills
Skills for the paid Scenario service: images, video, audio and 3D from an agent, model and price selection before spending, sets for game assets and brands
A set of SKILL.md skills for short-video and content production: topic, script, voice, avatar, editing, covers and motion graphics
README in short
The repository is set up to hold several Blender skills; today it contains one. Skills install individually or all at once, globally or per project. Work follows 11 phases with a gate on each.
README badge
Original SKILL.md
--- name: blender-image-to-3d description: Build a game-ready 3D asset in Blender from reference images (concept art, photos, turnarounds, sketches, screenshots) for any category, including characters, creatures, architecture, vehicles, props, weapons and environment pieces, through gated phases with rendered evidence compared against the reference. Use this whenever a user supplies or mentions a reference image and wants a 3D model or game asset made from it (modelled, blocked out, sculpted, retopologised, textured, rigged, animated or exported for a game engine), even when Blender is not named. Includes bpy scripts for scene calibration, review renders matched to the reference camera, silhouette measurement, validation, baking, GLB/FBX export with manifests, and clean reimport. license: MIT --- # Blender image to 3D Turn one or more reference images into a delivered game asset by moving through gated phases. Every gate is passed with rendered evidence compared to the reference from the same camera and at gameplay pixel size, never by looking at the viewport and deciding it seems fine. Most wasted work in this pipeline is detail added on top of wrong proportions, so nothing gets detail until its silhouette passes. "Perfect" here has a definition: the acceptance checklist in `references/delivery-and-acceptance.md` passes, the compare sheets show no measured deviation above the phase tolerance, and every dimension or side that the reference did not show is listed as inferred. Say so when a single image forces inference; do not present a guessed back as fact. ## Runtime - Blender 4.2 or newer (used through 5.2), run headless: `blender --background --python <script> -- <args>`. Blender 5.x changed several APIs these builds touch; `references/blender-5-notes.md` lists them with runtime, baking and review lessons. Resolve the binary once: `$BLENDER_BIN`, then `blender` on PATH, then `/Applications/Blender.app/Contents/MacOS/Blender`. Workbench renders need no GPU on macOS or Windows; on a headless Linux box without a GPU, pass `--engine cycles` to review_render.py. - A live Blender session through an MCP server (execute-Python tool) is optional and useful for inspection. Even then, build through the numbered scripts and save the same master file, so the build stays reproducible. In a live session, `exec(open(path).read())` a script and call its functions instead of the command line. - All modelling is written as bpy code in `<asset>/build/NN_<phase>.py`, copied from `assets/build_template.py`. Each script opens the master, deletes what it owns, rebuilds it, saves. Rerunning any phase is safe. Constants are measured metres with a comment naming the reference view they came from, or `# inferred`. - Skill scripts (all take `-- --help`): | Script | Purpose | | --- | --- | | `scripts/init_master.py` | master .blend: units, collections, calibration proxies, reference planes at real scale, game camera | | `scripts/review_render.py` | clay / silhouette / wire / checker / material renders from fixed views, the reference-matched camera, gameplay pixel size, turntables, posed frames | | `scripts/compose_review.py` | compare sheet (reference, render, overlay, gameplay strip) plus silhouette IoU and width-profile numbers; plain Python with Pillow | | `scripts/world_gate.py` | world-registered silhouette IoU for orthographic reference views: the camera covers the reference matte's exact world window, so scale and placement errors count | | `scripts/validate.py` | topology, transforms, UVs, weights, armature, sockets, colliders, naming, tri budget; exit 1 on FAIL | | `scripts/bake_maps.py` | normal and AO from HIGH to LOW, base colour, roughness, metallic; rebuilds delivery materials |