Layer Textures

layerai/skills/skills/layer-textures

by layerai315d06db6f76MIT4 starsListed Oct 8, 2026Updated Oct 8, 2026Repository updated 8 days ago

Use when generating a texture with Layer that must repeat without a visible seam: ground, terrain, walls, fabric, foliage, stone, wood, pattern fills, or a tiling material for a game engine. Also when a generated texture shows seams at its edges, an obvious repeating motif across a surface, or baked-in lighting that fights the engine. Keywords: tileable, seamless, tiling, texture, material, pattern, trim sheet, terrain, repeat.

Instructions onlyDesign & Creative
AI-generated overview

Guides generation of seamless, tileable textures with the Layer image model, covering scale, lighting and repeat checks.

What it does
This skill gives instructions for producing textures that repeat without visible seams, such as ground, stone, wood, fabric or terrain materials. It covers selecting a model with tileability capability, stating real-world scale, requesting flat diffuse lighting and uniform variation, and judging results at a 3x3 repeat. It also advises on obtaining PBR or material outputs in the same run rather than deriving normal maps afterwards.
When to use it
Use it when generating a texture or material that must tile seamlessly for a game engine or surface fill. It also applies when an existing generated texture shows edge seams, an obvious repeating motif, or baked-in lighting.
Requirements
Requires the Layer image-generation tooling and the sibling layer skill for generation mechanics; the skill is instructions only and ships no scripts. It references list_base_models, get_base_model and batch execution, and may require installing sibling skills such as layer-pixel-art.

Layer Textures

Overview

A texture is not an image of a surface. It is an image whose left edge continues into its right edge and whose top continues into its bottom, so that a plane tiled with it shows no seam. A model will not do this unless the run asks for it.

The mandatory step is the capability filter: list_base_models with filter.capabilities: {tileability: true}. Note the shape: capabilities is an object of booleans, not a list of names. A model without it produces a picture of stone that seams visibly the moment it repeats, and no editing pass fixes that afterwards. Everything else on this page is secondary to getting that filter right.

Generation mechanics are in the layer skill. If a sibling skill named here is missing from your available skills, ask the user to install it (npx skills add layerai/skills --skill <name>); unattended, proceed from tool schemas and flag the gap.

State the scale

Scale is what separates a usable texture from an unusable one, and it is the most commonly omitted field. Say how much real-world surface the tile covers: "a 2-metre square of cobblestone, roughly twenty stones across" is a texture. "Cobblestone" could be one boulder or a city square, and the model will pick.

Anchor it to the feature rather than the tile: "individual planks about 15cm wide", "grass blades roughly 5cm", "brick courses at standard size". The engine will scale the material, but only if the density inside the tile is right to begin with.

Flat light, no hero features

Two prompt clauses do most of the remaining work:

  • Even, diffuse, flat lighting, no directional shadows. Baked-in shadow is the second-biggest defect after seams. A texture lit from the upper left looks wrong everywhere the engine's light is not upper left, and it cannot be removed later.
  • Uniform distribution, no focal point, no single distinguishing feature. Generative models like to compose, so they will place one dramatic cracked stone in the middle. Tiled across a floor, that crack becomes a grid of identical cracks and is the thing every player notices.

Ask for consistent variation instead: "evenly distributed wear and colour variation across the whole tile, no single dominant feature."

Judging the result

Do not judge a texture by looking at the tile. Judge it tiled. Check the four edges continue into each other, then check a 3x3 repeat for a motif that has become a pattern. A tile that looks excellent alone and shows an obvious rhythm at 3x3 is a failed tile.

If the seam is right but a feature repeats too visibly, the fix is usually a rerun with stronger uniformity language rather than an edit, since removing the feature locally reintroduces a seam.

Material maps

A texture for a lit engine scene usually needs more than colour. Where a model declares PBR or material outputs, take them in the same run rather than deriving normals from the colour map afterwards, which produces flat-looking surfaces. get_base_model is the authority on what a given model returns.

Pixel-art tiles carry these rules plus their own: see layer-pixel-art.

Worked example

"A seamless mossy stone floor for our dungeon."

  1. list_base_models with filter.use_case: "text_to_image" and filter.capabilities: {tileability: true, square: true}. Take the first result. Without tileability this run is wasted.
  2. get_base_model, to see whether it exposes a tiling parameter that must also be set, and what material outputs it offers.
  3. Prompt: "Seamless tileable texture of worn dungeon flagstones with moss in the joints, covering roughly 2 metres square with about eight stones across. Even diffuse flat lighting, no directional shadows, no cast shadows. Uniform distribution of wear and moss across the whole tile, no focal point, no single dominant crack or feature. Top-down orthographic view."
  4. Estimate, then execute with batch_size: 4 so there are options.
  5. Judge each candidate at a 3x3 repeat before choosing.

Common mistakes

  • Skipping the tileability filter, which no later step can compensate for.
  • Omitting scale, so the tile is one enormous stone or a thousand tiny ones.
  • Accepting directional lighting baked into the tile.
  • A dramatic hero feature that becomes a visible grid when repeated.
  • Judging the tile on its own rather than tiled at 3x3.
  • Deriving normal maps from the colour map when the model could return them directly.
  • Requesting a perspective view of a surface instead of a top-down orthographic one.

Source and attribution

Source:layerai/skillsinskills/layer-texturesat commit315d06d

License: MIT

Content belongs to its original authors. SourceWeft indexes it from a public repository.

Report or request removal

More from layerai/skills

Layer Workflows

layerai

Use when running a saved Layer Blueprint workflow rather than a single generation: discovering what workflows a workspace has, reading a workflow's input schema, estimating and executing a run, polling its steps, or cancelling it. Also when a repeatable multi-step pipeline exists for a task, when the user names a workflow, or when a workflow input expects a style or a file. Keywords: workflow, blueprint, pipeline, app, multi-step, run, node graph.

Awaiting classification4updated 8 days ago

Layer Workflow Import

layerai

Rebuilds workflows from other tools as Layer Blueprint graphs and imports them via import_workflow.

AI & Agents4updated 8 days ago

Layer Video Timeline

layerai

Assembles existing clips, images and audio into one finished video timeline with transitions, overlays and captions.

Design & Creative4updated 8 days ago

Layer Video

layerai

Use when generating video with Layer: text-to-video, animating a still image, extending a clip, adding camera motion, generating native audio or lip sync, looping animations, or planning a multi-shot ad, trailer, or cutscene. Also when a video prompt produces the wrong motion or the shot drifts off the source image. Keywords: txt2vid, img2vid, image to video, camera motion, video effects, loop, seamless, trailer, cutscene, lipsync.

Awaiting classification4updated 8 days ago

Layer Reference Sets

layerai

Use when a look must hold across many Layer generations: training a custom style or LoRA on a studio's own artwork, curating the images that go into a reference set, choosing its training category, tuning reference-set weight on a run, or deciding whether to train at all rather than attach a style reference. Also when a trained style produces weak, inconsistent, or silently ignored results. Keywords: LoRA, custom model, trained style, reference set, consistency, on-model, art direction, dataset.

Awaiting classification4updated 8 days ago

Layer Quality

layerai

Reviews generated asset batches against a studio's output scoring rules and the brief before delivery.

Design & Creative4updated 8 days ago