Castaly · Blog · Pricing & billing

How Castaly Billing Works: Estimates, Holds, Refunds and Your Balances

What happens between Generate and the final charge: the estimate, the hold, settlement and refunds, plus how paid and gift balances are shared and spent.

Paper documents on an office desk at night curl up into a breaking wave under a desk lamp, an image made with GPT Image 2.5 Flare.

Castaly bills each generation from a balance you top up in advance. No subscription, no credits to convert, and no minimum. But between pressing Generate and seeing the final charge there are a few steps worth understanding: an estimate, a hold, a settlement, and sometimes a refund. And your account can hold up to three different balances.

This guide explains all of it with real examples.

Your three balances

The Top up page shows up to three amounts:

Balance What it pays for Notes
Paid balance Castaly and the Modelflare API Never expires
Castaly gift balance Castaly only Spent first in Castaly
Modelflare gift balance The Modelflare API only Spent first on API calls

Castaly runs on your Modelflare account, so there is one paid balance for both. A top-up made on either site lands in the same place, and you never move money between products.

Gift balances never mix. In Castaly, your Castaly gift balance is spent first and your paid balance covers the rest. Modelflare gift balance cannot be spent in Castaly, and Castaly gift balance cannot pay for API calls.

Gift balance comes from promotions. For example, when a friend registers with your invite link and completes a first eligible top-up, you each receive 5% of that amount as Castaly gift balance, 24 hours later.

From Generate to final charge

1. The estimate

Before you generate, the Studio shows an Estimated cost for the model and settings you picked. For per-second models it is the rate times the length. For token-billed models it is a budget for that size, quality or length.

2. The hold

When the task starts, Castaly reserves the estimate from your balances: Castaly gift balance first, then paid balance. If the two together cannot cover it, the task does not start. Nothing has been charged yet; the money is set aside.

3. Settlement

When the result is ready, Castaly works out the final charge from what the provider reports: the length of a finished clip for per-second models, and the tokens used for token-billed ones. If the final charge is lower than the hold, the difference goes straight back. If it is higher, the difference is collected then.

4. Refunds for failures

A task that fails is refunded in full. The part of the hold that came from your Castaly gift balance goes back to the gift balance, and the rest goes back to your paid balance.

How far apart estimate and charge can be

For per-second video models, the estimate and the final charge usually match closely, because the rate is fixed and only the reported length can differ.

Token-billed models can be further apart, in your favour. The Studio holds enough for the full budget and settles on the tokens actually used:

Generation Estimate held Settled on reported usage
GPT Image 2.5 Flare, 1536 × 1024, medium $0.044 about $0.011 (343 output tokens)
GPT Image 2.5 Flare, 1536 × 1024, quality not set $0.044 about $0.020 (601 output tokens)
Gemini 3.1 Flash Image, 2K $0.170 about $0.106 (1,680 image + 411 text tokens)
Paper documents on an office desk at night curl up into a breaking wave under a desk lamp, city towers lit outside the window.
gpt-image-2.5-flare · 1536 × 1024 — 601 output tokens and 99 input tokens reported, about $0.020 at today's price.

The largest holds come from Auto duration on video. It lets the model choose the length, so the hold covers the longest clip it could return: about $8.63 for Seedance 2.5 at 720p, even if the clip comes back at six seconds and settles at around $1.72. Choose a fixed length when your balance is tight.

A worked example

Say you have $1.00 of Castaly gift balance and $10.00 of paid balance, and you start a 5-second 720p Seedance 2.5 clip estimated at $1.44.

  1. Hold: $1.00 from the gift balance and $0.44 from the paid balance. You now have $0 gift and $9.56 paid available.
  2. Settlement: the clip finishes at the expected size, so the final charge is $1.44 and nothing changes.
  3. If it had failed: $1.00 would return to your Castaly gift balance and $0.44 to your paid balance.

Where to check your spending

  • Creations lists every generation. Open one to see its Cost summary: the Final total and any Adjustments and refunds.
  • Top up shows your paid and gift balances, and your top-up history.
  • The pricing page lists every model's public price, the same price the API charges.

Refunding a top-up

An eligible top-up can be refunded from your top-up history. The unused paid balance from that top-up goes back to the original payment method minus the processing fee, and any bonus credit that came with it is removed.

FAQ

Do I need a subscription?

No. You pay per generation from your balance, and paid balance never expires.

Why was less charged than the Studio estimated?

For token-billed models the estimate is a hold sized for the full budget. The final charge uses the usage the provider reports, and the difference is returned automatically.

Can I use my Castaly gift balance through the API?

No. Castaly gift balance only pays for generations in Castaly. API calls use your Modelflare gift balance first, then your paid balance.

Is a failed generation charged?

No. A failed task is refunded in full, including any gift balance it used.

Is the price in Castaly different from the API price?

No. Castaly uses the same public price list as the Modelflare API.