Game dev sprint: shipping a vertical slice in 30 days with AI 3D

One of the most asked questions in our Discord is some version of: “Can I actually ship a real game with AI 3D, or is it just for demos?” Here's a real-world answer. (For the higher-level math behind the question, see AI 3D vs traditional pipeline.)

The team

A 3-person studio in Helsinki: one programmer, one art director, one game designer. No dedicated 3D artist. Their goal: a 30-minute vertical slice of a stylised dungeon-crawler RPG, ready to show publishers.

The brief

• ~50 unique 3D assets needed (player, 12 enemies, environment kit, weapons, props)
• Stylised lowpoly art direction — see our lowpoly vs high-detail guide for the trade-offs.
• Mobile target (Unity, polycount budget ≤8K per character)
• 30 days to playable

Day-by-day breakdown

Days 1–3 — Pre-production

Art director sketched the visual language. Mood board, palette, style references. Five hero characters were generated in Polyx with iteration on prompts until silhouettes felt right. Lowpoly mode + A-pose for clean rigging downstream.

Days 4–10 — Character production

The remaining 7 enemies generated in batches. Each character: prompt → preview → refine → auto-rig → animation library applied (idle / walk / attack / die). Average time per character: 6 minutes including iteration.

Art director did manual cleanup on 3 characters with extra silhouette tweaks. The other 9 went straight to Unity.

Days 11–18 — Environment kit

Modular dungeon kit: walls, floors, doors, columns, debris. Generated as standalone props in lowpoly mode. ~80 unique meshes total. Most were generated, the rest were retextures of existing meshes via AI Texturing for variation.

Days 19–24 — Props and weapons

Swords, shields, potions, treasure chests. Smaller batches, faster iteration. Programmer started gameplay integration in parallel.

Days 25–30 — Polish and ship

Animation tuning, lighting, UI integration, bug-fixing. 30-minute playable slice ready by day 30.

The numbers

• Total 3D assets: 142 (50 unique + variants)
• Polyx credits spent: ~1,800 (~$240)
• Manual touch-up time: ~25 hours total across all assets
• Equivalent freelance budget: $18,000–$25,000
• Equivalent timeline: 3–4 months

What worked

• Generating in batches. Iterate prompts on 5 characters at once, pick the best 2.
• Lowpoly mode + auto-rig. Cleanest pipeline for game-ready characters.
• AI Texturing for variation. Same enemy mesh, different material — see how PBR maps work.

What didn't

• Highly bespoke armour rigs needed manual weight-painting. Auto-rig deformations were “good not great” on chunky armour shapes.
• Environment lighting needed lots of tuning — generated assets exposed lighting setup choices that hand-authored assets would have hidden.

Verdict

The team got their publisher meetings. Two offers. They've signed.

AI 3D doesn't replace senior tech artists. But it lets a 3-person team ship what used to need 8.

Continue reading

• Polyx vs the traditional 3D pipeline — the math behind the speedup.
• Lowpoly vs high-detail — picking the right mode for mobile.
• Auto-rigging in 60 seconds — the rigging side of this sprint.

Ready to start your own sprint? Open the workspace →