2D Top-Down Tileset & Prop Artist (AI-Generated) - Virtual Tabletop Map Builder

About the role RollQuest is a virtual tabletop (VTT) platform for tabletop role-playing games. We've just built a Map Builder - an in-app tool where users stamp tiles, paint floors, draw corridors and rooms, and drop in props to create dungeon battlemaps on a snapping grid. It currently ships with placeholder art. We're looking for a 2D artist to produce a cohesive, production-quality top-down dungeon tileset and prop set to replace it. The art must drop into our existing engine with no code changes - which means it has to hit a precise technical spec so tiles tile seamlessly and walls/corridors connect cleanly. This is a well-scoped first project (~40 assets) with strong potential for follow-on themed packs (cave, crypt, forest, sci-fi, etc.) if it goes well. This is an AI-generation role. We require the assets to be produced with AI image tools - we're optimising for speed and volume, not hand-painting from scratch. The single most important requirement is consistency: every tile and prop must read as one cohesive set (same palette, lighting, line weight and detail level) so they link up seamlessly. A fast, repeatable AI workflow that holds a locked style across the whole set is exactly what we're after. What you'll do - Produce a complete classic stone dungeon tileset: floors, walls, corridors (including a 16-piece auto-tiling set), doorways, stairs, pillars, and ~12 furniture/clutter props - Deliver everything to our exact grid spec so it slots straight into the Map Builder - Establish a consistent palette, lighting model, and level of detail across the whole set - Provide a short style sheet so future packs stay consistent - Iterate with us on a small first batch before producing the full set The technical spec (this is what makes it "link up") Please read this carefully — meeting it is the core of the job. - Grid unit = 100px per square. Each asset spans a whole number of squares (its "footprint"). Its image **aspect ratio must match its footprint** (a 2×3 prop = 2:3). Author high-res, deliver at the target sizes below. - True top-down orthographic Straight bird's-eye view, no perspective, no isometric, no visible sides. Flat, even ambient lighting with soft self-shadow only - no long directional cast shadows (assets get rotated and flipped in-app, so directional shadows break). - Floors must be seamlessly tileable on all four edges (they repeat across many cells with no visible seam). - Walls are thin and edge-aligned - a consistent ~12% of a cell thick, hugging the cell boundary, identical thickness on rooms, corridors and wall segments, so adjacent pieces form continuous lines. - Auto-tiling matters. The 16-piece corridor set must be edge-matched so any two pieces connect: every open edge presents identical floor, every closed edge presents an identical wall. (Wang/blob/auto-tile experience is essential here.) - Transparency rules: floors and full-cell corridor pieces are opaque; props, wall segments, pillars and doorways are transparent PNG (or WebP) with clean alpha so they sit over floor. - One cohesive palette and light model across the entire set. The asset list (~40 pieces) We'll provide exact filenames so delivery is drop-in: - Floors — seamless, 2 variants (1×1 and 2×2) - Corridors — straight (H/V), corner, cross, plus a 16-tile auto-tiling set (1×1 each) - Walls & structure — horizontal/vertical wall segments, doorway, pillar, stairs - Rooms — 4×4 and 8×8 floor+wall tiles - Props (~12) — barrel, crate, chest, table, chair, bed, bookshelf, brazier, rug, statue, fountain, rubble (footprints range 1×1 to 3×2) Deliverables - All assets as PNG/WebP with correct transparency, named to our convention (we'll supply the list) - Layered source files (PSD / Procreate / Affinity / SVG — whatever you work in) - A one-page style sheet (palette, lighting, wall thickness, conventions) so future packs match Workflow — AI-generated, consistent, fast AI generation is required, not optional. We care about three things, in order: consistency (the whole set looks like one pack), correctness (hits the technical spec so it links up), and speed (an efficient, repeatable pipeline). Show us you can lock a style - fixed seeds/style references, consistent model/LoRA, a master style anchor - and stamp out a coherent 40-piece set quickly. To make consistency even easier, we have a compositing generator on our side: if you prefer, you can deliver just the seamless base materials** (one floor, one wall stone, plus optional wood/metal/water/cloth swatches) and the standalone props, and we'll assemble the corridor auto-tiles, room walls, doorways and stairs from your materials. Because everything is built from the same two materials, this route guarantees they tile and stay consistent. We're happy either way - full set, or materials-plus-props we composite - tell us which you'd prefer in your proposal. Must-have skills - Strong AI image-generation skills - fluent with at least one pro tool (Midjourney, Stable Diffusion / ComfyUI, Flux, etc.) and, crucially, the techniques for a consistent set: locked seeds and style references, consistent model/LoRA, controllable seamless + transparent output - Proven ability to keep a large set visually consistent — same palette, lighting, line weight and detail level across every asset, so they read as one pack and link up - Experience producing seamless/tileable textures and auto-tiling / Wang / blob tilesets (the make-or-break technical skill) - Clean transparent sprite/prop work with proper alpha - A fast, repeatable pipeline — you can turn the full set around quickly and re-run it for revisions/new packs - Comfortable working to a precise technical spec and iterating from feedback Nice-to-have - Familiarity with VTTs / tabletop RPGs (D&D 5e, Roll20, Foundry, Dungeondraft, etc.) - Experience authoring assets for grid-based / auto-tiling engines - Ability to deliver additional themed packs later (cave, crypt, forest, sci-fi) To apply, please include 1. A portfolio link with top-down tile/battlemap work. If you have seamless/tileable or auto-tile (Wang/blob) examples, call them out specifically - that's the make-or-break skill for this job. 2. A short note on which AI tools/workflow you'd use, and crucially, how you keep a large set visually consistent (locked seeds, style references, consistent model/LoRA, master style anchor) so every tile reads as one pack. 3. Whether you'd prefer to deliver the full tileset, or base materials + standalone props for us to composite. 4. Your availability and rate.