Separate outcome from cost
Pixelcut documents a credit-based API: successful requests consume credits, while failed requests typically do not. “Typically” matters. The documentation also identifies safety moderation as an example where credits may still be consumed. That means an error response is not enough evidence to justify an automatic retry.
Use two independent fields for every request: operational outcome and credit state. Operational outcome answers whether a usable result arrived. Credit state answers whether the balance has been confirmed, inferred from the documented general rule, or remains uncertain. Do not collapse both into a single “failed, retry” flag.
- Usable result: stop generation and move to asset review.
- Ordinary failure with no moderation indication: a retry may be considered, but record that the no-charge rule is only typical.
- Safety or moderation-related failure: treat the charge as uncertain until the balance is checked.
- Unknown response, timeout, or interrupted client: do not immediately replay the same request; first reconcile the request record and balance.
Use a balance checkpoint
The official route for a programmatic balance check is GET /v1/credits, authenticated with the X-API-Key header. Pixelcut also says usage can be monitored through the developer dashboard. The source does not specify a response schema, operation-by-operation price list, or a guaranteed delay between a request and reflected balance, so do not invent a subtraction formula from undocumented fields.
A practical planning aid is a checkpoint ledger. Before a batch, record the observed balance and the intended request identifiers. After each completed request, record the HTTP outcome, whether an output was received, any moderation indication, and the next observed balance when available. This gives the operator a basis for deciding whether to continue, pause, or investigate without pretending the ledger proves an exact charge.
- Pause when the post-request balance is unavailable and the next request would duplicate an uncertain operation.
- Continue only when the request has a usable result or the team has explicitly accepted the unresolved credit risk.
- Escalate discrepancies with timestamps, request identifiers, response status, error body, and balance observations.
Define the retry boundary
A retry policy should specify what counts as the same work. Repeating an identical request after a timeout may create a duplicate output or consume another credit; changing the prompt or source makes it a new creative attempt and should be tracked separately. The supplied documentation establishes credit behavior at a general level, not idempotency or duplicate-request protection.
For a production workflow, make the next action depend on evidence: retry ordinary, clearly received failures only within your internal retry limit; hold moderation-related failures for review; and reconcile ambiguous transport failures before replaying. Keep a human approval step for a batch when the remaining balance cannot cover even one additional uncertain request.
Hypothetical batch decision
Hypothetical example only: a seller has an observed Pixelcut balance of 12 credits and plans five product-image requests. Request A returns a usable image, so the team records success and proceeds to review. Request B returns an ordinary error; the team records the failure and considers one bounded retry. Request C returns a moderation-related error; the team does not replay it until a new balance observation is available. Request D times out after submission; because the server-side outcome is unknown, the team pauses rather than assuming the request was free.
The result is not a calculated remaining balance, because the source supplies no universal operation price or response schema here. The actionable decision is narrower: A can move to review, B can enter a controlled retry path, C requires moderation review and reconciliation, and D requires outcome evidence before duplication.
| Request state | Credit assumption | Next action |
|---|---|---|
| Usable output received | A successful request may consume credits | Keep the result and record the observed balance |
| Ordinary error | Failed requests typically do not consume credits | Consider one bounded retry after recording evidence |
| Safety or moderation error | Credits may still be consumed | Check balance and review before replaying |
| Timeout or unknown outcome | Charge and server outcome are uncertain | Reconcile first; do not automatically duplicate |
Related questions
Does Pixelcut document the exact credit cost of every API operation here?
No. This page explains when credits are generally consumed and how to check usage, but it does not provide a universal per-operation price table. Confirm the current operation-specific documentation or dashboard information before budgeting a batch.
Where CartVisual fits
After obtaining an image from any provider, use CartVisual’s browser-local checklist to compare the result with the approved product source. CartVisual does not expose Pixelcut’s API balance or execute Pixelcut retries.
Source and evidence scope
Prepared by the CartVisual editorial team with AI assistance. Competitor behavior is documented rather than independently executed for this guide.
Official source snapshot checked .
Pixelcut’s official Credits documentation, checked 2026-11-17. It covers general consumption behavior, the balance endpoint, and dashboard monitoring. It does not establish operation prices, response schemas, idempotency, timing guarantees, or automatic retry behavior.