How Studio AI model access, BYO providers, and managed credits are intended to work.
Providers & AI Credits
Studio AI can run against different provider and billing setups depending on your plan, deployment, and workspace configuration.
The core principle is simple: heavy AI usage should be controllable, predictable, and visible.
Two Operating Models#
Studio AI is designed around two possible ways to pay for model usage.
1. BYO provider mode#
In BYO mode, you connect your own model providers for text or media generation.
This is the preferred long-term operating model for power users because:
- It scales with your own provider account.
- It avoids surprises from a shared platform credit pool.
- It fits teams that already standardize on one model vendor.
2. Managed credit mode#
In managed-credit mode, ED5 fronts the provider call and charges usage against your Studio AI credit allowance or purchased top-up balance.
This mode is useful for:
- onboarding and trial experiences
- light convenience usage
- users who want one-click AI without connecting providers immediately
Credit Policy Basics#
Depending on your plan and configuration, these rules should hold:
- BYO-provider usage should not consume ED5-managed credits.
- Managed-provider usage can reserve credits before the provider call starts.
- Failed generations should refund reserved credits when no usable output is created.
- Different models and quality settings can consume different amounts.
- Batch jobs and media generation usually cost more than short read-only text runs.
What Usually Affects Cost#
The biggest cost drivers are:
- Large image generations or multiple candidate sets.
- High-end text models.
- Long multi-step runs that call many tools.
- Repeated iteration without narrowing the brief.
- Broad validation or simulation work across many records.
If you want lower cost, constrain the run:
- ask for a plan first
- limit the number of generated variants
- target a single zone, item family, or quest arc
- reuse existing project references instead of asking for wide exploration
Example Cost Patterns#
| Task shape | Typical cost profile |
|---|---|
| Explain how a system works | Low |
| Review one quest or NPC set | Low to medium |
| Create a small batch of items | Medium |
| Generate several icon or portrait variants | Medium to high |
| Batch worldbuilding plus media plus validation | High |
Recommended Setup Strategy#
If you are just evaluating Studio AI:
- Start with small read-only or single-entity tasks.
- Use managed credits if your plan includes them and the workspace enables them.
- Measure how often you actually need media generation versus text planning.
If you are using AI heavily in production:
- Prefer BYO providers where available.
- Keep managed credits for convenience or occasional overflow.
- Define approval and autopilot rules that match your spend tolerance.
Security Notes#
Provider credentials should be handled server-side and never exposed directly to browser JavaScript after creation.
Developer API keys for Creator MCP are not provider keys. They only authorize access to the Studio AI tool surface.
Troubleshooting#
| Problem | What to check |
|---|---|
| AI run is blocked for billing or credits | Confirm whether the workspace expects BYO providers or managed credits |
| Credits seem to drop too quickly | Check model choice, image size, batch size, and how many retries or variants were requested |
| BYO usage still appears billable | Verify whether the current run actually used your connected provider or a managed provider path |
| Media run failed but credits stayed reserved | Retry after the run closes, then review whether the output was marked unusable and refunded |