Asset Pipeline logo

Asset Pipeline

CommunityPopular
rehan-remade
asset-pipeline

Turn generated or hand-made art into exactly what a game engine loads. Covers sprite cutout, trim and nearest-neighbour fitting to frame sizes, pixelation to a palette, sprite sheets and strips, player/team-colour masks, seamless textures, and rendering a 3D model into sprite frames from the game's own camera (isometric/RTS 8 or 16 headings, side view, top-down) with Blender. Use after generating assets, or when the user asks to make sprites, sprite sheets, animation frames, isometric unit sprites or icons for a game.

Overview

Publisherrehan-remade
Repositoryuniversal-modder
Skill nameasset-pipeline
Stars
3.4K
Forks
293
Bundled files
Instructions only
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.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by rehan-remade on GitHub. Read the source before you install it.

Installation

Install the Asset Pipeline 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/rehan-remade/universal-modder/tree/main/.agents/skills/asset-pipeline
  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/rehan-remade/universal-modder.git /tmp/universal-modder
mkdir -p .claude/skills
cp -r /tmp/universal-modder/.agents/skills/asset-pipeline .claude/skills/asset-pipeline
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Asset Pipeline 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 Asset Pipeline 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 Asset Pipeline 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.

Asset pipeline: from art to engine-ready files

The fal-assets skill makes pictures. This skill makes files the game accepts: the right size, frame layout, orientation, alpha, palette and format. All commands are um sprite ... and um render3d ...; each has --help with examples.

1. Learn the target format from the game itself

Before converting anything, open two or three of the game's own assets and write down in MODLOG.md:

  • Dimensions: frame size, frames per sheet, and the layout (vertical strip / grid / one file per frame).
  • Orientation: which way the art faces (Terraria: items point right, NPCs face left, the engine flips them), and where the pivot or hotspot sits (AoE2: the unit's ground point at the canvas centre).
  • Alpha: hard 1-bit edges (BC1 punch-through, pixel art) or soft.
  • Style: palette size and outline (dark 1-2 px outline in most pixel-art games).
  • Format: PNG, DDS BC1/BC3/BC7, XNB, SLD, atlas + JSON...

The draw code is the final authority. In Terraria, NPC frame height = texture height / Main.npcFrameCount, so any consistent frame size works.

2. 2D pipeline

bash
um sprite info raw.png                                    # size, alpha coverage, corner colour
um sprite cutout raw.png cut.png                          # flat bg → transparent (border flood fill keeps interior whites)
um sprite cutout raw.png cut.png --grey 150 --keep-top 0.8   # also remove a soft grey shadow / the floor under it
um sprite fit cut.png item.png --size 64x26 --hard-alpha  # trim + ONE nearest-neighbour scale into the frame
um sprite fit cut.png npc.png --size 38x34 --anchor bottom   # standing sprites sit on the frame's bottom edge
um sprite pixelate cut.png px.png --size 32x32 --colors 16 --outline   # true pixel art from painterly art
um sprite palette px.png px2.png --from stock_sprite.png  # snap to the game's own colours
um sprite frames npc.png frames/ --n 3 --kind bob          # cheap idle animation (bob/squash/wobble/flash)
um sprite sheet sheet.png frames/*.png --vertical         # Terraria-style strip (or --cols N grid)
um sprite slice sheet.png out/ --frame 32x32              # the other direction
um sprite team-mask unit.png unit_grey.png --hue blue     # player-colour mask (+ desaturated sprite)
um sprite preview item.png look.png --scale 6             # checkerboard + zoom: LOOK at it before shipping
  • Pixel art: scale once, with nearest neighbour, to the final size. Never scale pixel art twice.
  • Painted / HD art: use fit --smooth.
  • Real frames: for animation frames beyond bob/squash, generate each frame with the fal edit endpoint using the base sprite as reference ("same drone, rotors tilted, frame 2 of 4"), then cut out and fit each frame the same way. Or go 3D (below).

3. 3D → sprites (consistent angles and animations)

bash
um fal model3d concept.png --name unit                   # textured GLB (Trellis 2 by default)
um render3d assets/gen/unit.glb frames/ --preset aoe2 --length 80 --forward-yaw -90 \
  --anims idle:10:bob,walk:12:walk,attack:16:lunge,death:20:die --shadows --samples 40
um render3d assets/gen/unit.glb side/ --preset side --canvas 128 --length 110 --engine eevee   # platformer facing R + L

Presets:

PresetProjectionCameraFacings
aoe2ortho30°16 clockwise from east
iso8ortho30°8
trueisoortho35.264°8
topdownortho—8
sideortho—2 (right, left)
turntablepersp—24 (promo/icon spins)
  • --length sets the model's longest horizontal side in pixels at 1x, so match stock units.
  • --forward-yaw turns the model so its nose faces +X; check the first frame.
  • --shadows adds a shadow-only pass (*_s.png) for engines that keep shadows in their own layer.
  • Motions: bob, walk, lunge, die, wreck, spin.
  • Then pack with um sprite sheet, or with an engine writer (e.g. examples/aoe2-de-civ/sld.py).
  • Needs Blender (blender on PATH or BLENDER=...). Cycles uses the GPU when available.
  • Dark generated textures: raise --sun / --ambient, or brighten in post.
  • For game-ready 3D (not sprites), remesh with um fal run tripo3d/tripo/remesh mesh_url=@unit.glb face_limit:=8000, then convert in Blender (GLB → FBX/OBJ) with the engine's scale and axis convention: Unity Y-up metres, Unreal Z-up centimetres, Bethesda NIF via PyNifly.

4. Textures and materials

  • Tiling: um fal texture generates it tiled. Check with um sprite tile-preview t.png t3.png. Fix seams on other images with um sprite seamless.
  • PBR sets: um fal pbr. Convert to the engine's packing: Unreal ORM (occlusion/roughness/metal in RGB), Unity metallic-smoothness (smoothness = 1 - roughness in alpha).
  • DDS: texconv (DirectXTex) or magick with DXT settings. BC7 for quality, BC1 for cutout sprites, BC3 when alpha is soft.

5. Verify in the game

Put one converted asset into the game and screenshot it next to stock art (um win shot). Check the scale, facing, pivot, outline and palette. Fix the recipe, then batch-convert the rest with the same commands (a small script or Makefile, so the pipeline is reproducible from assets/gen/).

6. Converting a guest game's Unity assets (mashups)

Read the user's own install with UnityPy and write a local, private resource pack. Never put the output in the mod, its repo or a release; each user runs the converter on their own copy.

  • Memory: UnityPy.load() decompresses a whole bundle (a 865 MB bundle became 8.7 GB of RAM). Decompress only the serialized files and read .resS / .resource lazily, block by block; run each converter module in its own process.
  • Environment: UnityPy.Environment(path=""), and load cross-referenced bundles together; otherwise UnityPy searches the working directory for missing dependencies.
  • Finding assets: Addressables bundle names are hashed per update (find them by prefix); Addressables 2.x has a binary catalog.bin. Map game objects to prefabs through their GUID keys, not by guessing names. Prefer the low-quality variant if one exists (same meshes, smaller textures).
  • Skinned meshes are stored in bind pose (often a T-pose). For the in-game look, evaluate the idle clip at t = 0 yourself (UnityPy reads AnimationClip objects but does not sample them; Generic clips bind Transforms by CRC32 of the path), rebuild world matrices, then skin.
  • Walk the hierarchy from the container root, skip inactive GameObjects and their children, and drop helper meshes (eyelids, ground shadows, stencil/outline materials).
  • Axes: Unity is left-handed, Minecraft right-handed: flip one axis for positions and normals and reverse the winding; flip V; clamp UVs. Verify with a named bone (the right hand ends on the right).
  • Shared atlases: crop each part's texture to its UV bounding box and remap UVs.
  • Audio: AudioClip.samples needs FMOD (fmod_toolkit); encode with ffmpeg to mono OGG.
  • Minecraft side: a pack needs pack.mcmeta (pack_format 15 for 1.20.1); ResourceLocations accept only [a-z0-9_./-]. For posed meshes, a custom mesh file drawn with RenderType.entityCutoutNoCull is more robust than forge:obj (block-atlas textures, usemtl).
  • Oracle: render preview sheets by re-reading the files you wrote, and look at them.

Frequently asked questions

What does the Asset Pipeline AI skill do?

Turn generated or hand-made art into exactly what a game engine loads. Covers sprite cutout, trim and nearest-neighbour fitting to frame sizes, pixelation to a palette, sprite sheets and strips, player/team-colour masks, seamless textures, and rendering a 3D model into sprite frames from the game's own camera (isometric/RTS 8 or 16 headings, side view, top-down) with Blender. Use after generating assets, or when the user asks to make sprites, sprite sheets, animation frames, isometric unit sprites or icons for a game.

Why use Asset Pipeline on TypingMind?

Because you install it once and use it with any model. Asset Pipeline 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 Asset Pipeline in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/rehan-remade/universal-modder/tree/main/.agents/skills/asset-pipeline. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Asset Pipeline?

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 Asset Pipeline?

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

Is the Asset Pipeline AI skill free?

Yes. It is published on GitHub by rehan-remade 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 👇