How a generation spends credits

From provider cost to credits

Here is the mechanism behind every charge. Each billable call Pinnora makes to a model has a raw provider cost in real dollars. Pinnora adds a retail markup to that cost, then debits the result as credits. At the current default, creative and media work (images, video, documents) is marked up 3x, so one credit covers roughly a third of a dollar of raw compute. Atlas chat turns and the App Builder run gentler, at 2x. The operator sets these multipliers and can change them, so read them as today's defaults, not permanent law.

This is why the same brief can cost different amounts on different days. A bigger batch, a longer video, a pricier model, or a research run that pulls more sources all raise the provider cost, and your credit debit climbs right along with it.

One writer, no double charges

Every credit movement flows through a single service, so the balance is never written from two places at once. Each charge carries an idempotency key tied to the exact thing it paid for, like one artifact or one run. It works like a numbered receipt: if a network blip triggers a retry, that key makes sure you pay once, not twice. Refunds and grants ride the same ledger, so nothing ever moves off the books.

See it happen

The fastest way to feel this is to run one job and watch. Open the social media ad generator, write a short brief, and dispatch a small batch. The balance in the top strip ticks down as each render lands, and the run shows up in your usage and ledger moments later.

Every concept in a batch is its own billed render, so eight ads is eight charges, not one. When you want to steer cost, the variants count on the options panel is your main lever.