Answers
Do I get my money back when an AI generation fails?
Short answer
Yes, automatically and in full. The charge happens when a generation starts and is reversed the moment it is marked failed — upstream errors, capacity errors and content-filter rejections all refund without a support ticket. A batch that is only partly rejected is refunded proportionally. There is nothing to claim and no waiting period.
- Failed generation
- Full refund to balance, automatic
- Content-filter rejection
- Full refund — counts as a failure
- Partially filtered batch
- Proportional refund for the rejected items
- Refund delay
- None — same transaction as the failure
- Paid over x402
- Images and speech never settle; video pays out a refund token
Facts on this page were checked against the live platform on .
When exactly am I charged?
At the start of the generation, not at the end. The balance check and the debit happen before the request leaves for the model, which is what stops two parallel jobs from overdrawing the same balance.
The consequence is that a running job shows as already paid for. If it fails, the reversal is written in the same code path that records the failure — the two are not separate steps that could get out of sync.
Which failures are refunded?
All of them. That includes model errors, capacity errors such as HIGH_LOAD, quota errors, expired upstream operations (VIDEO_EXPIRED), and content-filter rejections (SAFETY_FILTERED) — a filtered generation is a failed generation as far as billing is concerned.
Partial results are handled proportionally. Ask for four images, have one rejected by the output classifier, and you are refunded a quarter. The same rule applies when Veo returns some variants and filters others.
The one thing that is not refunded is a generation that succeeded and that you simply did not like. The compute was spent and the file is in your history — that is the risk that the cheap draft tiers exist to reduce.
How does this work for agents?
Over MCP it is identical, because agent calls go through the same billing code as the web studio: charged at start, refunded on failure, visible in the same usage log. An agent that retries a filtered prompt is not burning money, only time.
Over x402, where an agent pays per call in USDC without an account, the mechanics differ because a blockchain payment cannot be reversed. Images and speech are verified first and settled only after the file exists, so a failure never takes the money at all. Video has to settle up front — a running generation cannot be cancelled upstream — so a failed video answers with a refund token good for the full amount against a later call, valid for 90 days.
Where this is documented
Longer reads on the same thing
More on this question
Do I need to contact support for a refund?
No. Refunds are written by the system when the failure is recorded. If a balance ever looks wrong, the usage log in your profile shows every charge and reversal.
Are refunds paid back in cryptocurrency?
No — they return to your platform balance, which you spend on further generations. Deposits themselves are not reversed on-chain.
What if a video keeps failing?
You keep getting refunded, so repeated attempts cost only time. If a configuration fails four times in a row it is usually capacity: drop 4K to 1080p, which is priced the same as 720p on Veo 3.1 and Veo 3.1 Fast.
Does a refund token from x402 expire?
It is valid for 90 days and redeemed inside a later call; anything left over comes back as a new token for the remainder.
Questions next door
Why was my AI image blocked by the safety filter?
Because one of two independent filters fired. The first checks your prompt before anything is drawn and can be relaxed; the second inspects the finished pixels and cannot be turned off by anyone. A relaxed filter only affects the first. Since every run draws a different picture, a borderline prompt that fails the second stage often passes on a retry — and a blocked generation is refunded, so the retry costs only time.
Read the answerWhat does 'The service is currently experiencing high load' mean in Veo?
It is a capacity error from Google, raised inside a generation that already started — your prompt is fine. It hits hardest on the most expensive configurations, especially 4K with a last frame. The job is restarted automatically up to four times on the same key; if all attempts fail you get the error code HIGH_LOAD and a full refund.
Read the answerCan an AI agent generate images without an account or API key?
Yes, over x402. An agent holding a wallet posts a request, receives HTTP 402 with the exact price for those arguments, pays in USDC on Base and repeats the request with a payment header. No sign-up, no API key, no prepaid balance. Images and speech settle only after the file exists, so a failure costs nothing.
Read the answer