This is an experimental project. It does not sell tutorials or online courses. It is written for people who already run Cursor, Codex, or Claude Code every day and are looking for more capacity in the same routine.
What this actually costs
Subscription is the smaller line. Credit spend on generation is where the real cost sits, and it scales with how much art a project moves.
| Solo / side project | Small team | |
|---|---|---|
| Subscription | One account, medium credit allowance | One per account, medium credit allowance |
| Generation credits | Topped up as needed, roughly matching the subscription | Shared pool, administered by one manager |
| MCP access | Covered by the account | Billed per seat, separately |
| What the money buys | One person’s own build rhythm | Centralized control over shared assets |
Current spend on VberAI Studio here is about $60 a month: $30 of subscription and roughly $30 of credit top-ups. For a side project that has to fit around one person’s rhythm, and that pulls on several generation tools rather than one, that has been workable.
For a team the arithmetic changes. MCP access is billed per seat, each account carries its own subscription and credit allowance, and generation draws on a shared pool that one person administers. Current figures are on the VberAI site. The shared pool and the management tooling around it are the reason a team pays at all—a centralized studio, with one person accountable for the shared asset library, gets something from that structure which separate individual accounts do not.
This is not cheap. It fits a team in a rapid build-out phase, where throughput matters more than unit cost.
The rest of the stack
$100 a month was the planned AI budget for this project, across every service rather than VberAI alone:
- Cursor: $60
- T3: $8
- Figma: $15 (dev account, for MCP)
- VberAI: $30 subscription
With credit top-ups the real total runs closer to $130. Image generation goes through GPT Image 2 and Nano Banana 2 in t3.chat, design work through Figma and VberAI Studio, and code through Cursor.
What gets tested here
The working model in VberAI Studio is close to Figma’s: art sits on a canvas in layers, then moves to the engine through MCP—Godot MCP, Unity, and Cocos Creator. What this site records is how far that reaches in ordinary production work: where batch asset preparation holds up, and where generated art still needs another pass before it maps onto a scene.
Background:
- Designer, familiar with UI & UX
- Developer, familiar with the basics
- PM, familiar with product design, planning, and AI tools
- No background in game design or game development
Steps are documented with an “AI-first” mindset, at the grain an outsider to game development needs them.
If this exploration is useful, write to [email protected].