AI Image Generator — temporarily unavailable

Generation is under maintenance. The previous generation source is unavailable and its replacement requires authorized access. Valid runs currently return FAILED with SUMMARY.status=source_blocked and clear ERRORS. They deliver no images and create no image-generated events. The start fee and user-paid resource costs can still apply. Do not run a large generation batch while this notice is present.

The old title's “7 models” claim did not match the actual production backend. No seven-model selection, guaranteed fallback, generation speed, exact 4K, seed reproducibility or reference-image quality is currently verified. The Actor remains deprecated until useful generation is restored and tested.

Existing input contract

The current options are preserved to avoid losing existing workflows: prompts, negativePrompt, style, resolution, aspectRatio, numImages, seed, referenceImageUrl, saveToStore, compress and maxConcurrency. These are requested settings, not working-feature guarantees during maintenance.

Prompts must be a non-empty array of text strings, each up to4,000 characters. Maximum100 prompts in input; prompts × variants is capped at10 jobs and the excess is reported. Variants are the existing strings 1, 2, 4, 8; concurrency1–5. Style and aspect choices retain the input form's enums. Empty API input fails without an image result charge. Reference URLs must be public HTTP(S), without embedded credentials or private destinations.

Reports and costs

Dataset contains no error or placeholder image rows. Read SUMMARY and ERRORS rather than relying on an old SUCCEEDED status. ERRORS exposes prompt/variant indexes and safe error codes, without prompts, backend URLs or private routing settings. RETRY_INPUT is private run storage and may contain private prompts or reference links; do not publish it. Fix source access before retrying to avoid repeated start/resource charges.

SOURCE_ENDPOINT_UNAVAILABLE means the previous source route is absent. Its current replacement requires authorization. SOURCE_AUTH_REQUIRED, SOURCE_TIMEOUT, SOURCE_UNAVAILABLE and SOURCE_CONTRACT_REVALIDATION_REQUIRED require investigation; they are not proof that a prompt is invalid. Generation is intentionally blocked if an unexpected new contract appears, until useful output and billing are revalidated.

Current tariff: $0.004 per delivered image ($4/1,000) across tiers, plus start $0.00005 per GB, minimum one event, and separately user-paid compute/storage/proxy resources. The active resource flag is true, with no future pricing change. Failed or empty runs create no image events but are not zero-cost runs. Check the Pricing tab.

API and MCP diagnostics

Permanent Actor ID: Mp57DQfkhayPSf3TK; Store name: apivault_labs/ai-image-generator. This is a different Actor from ai-image-generator-nanabanana.

Example for a small diagnostic run after checking this notice:

{"prompts":["A blue ceramic teapot on a wooden table, no text"],"resolution":"1K","aspectRatio":"1:1","numImages":"1","maxConcurrency":1}

POST to https://api.apify.com/v2/acts/Mp57DQfkhayPSf3TK/runs using an Authorization Bearer Apify token. Poll the run ID, then read SUMMARY/ERRORS from defaultKeyValueStoreId and delivered rows from defaultDatasetId. Actor API. No token should be included in a public example, prompt or output URL.

Hosted Apify MCP tool: apivault_labs--ai-image-generator. Agent diagnostic request: “Run one small prompt only if I explicitly request a diagnostic despite the maintenance notice. Read Dataset, SUMMARY and ERRORS with get-dataset-items/get-key-value-store-record. Report that generation is blocked; do not claim an image exists from a successful API response alone.” API/MCP require Apify authentication. Automatic production generation should wait until the maintenance notice is removed after image QA.

Restoration requirements

Restoration requires authorized source access or a replacement agreed with the owner, followed by real generate/reference/style/negative-prompt/seed/variant/aspect/resolution/storage/compression checks, decoded image and visual review, delivered-image billing, production and MCP useful-result QA. This guarded error-reporting release does not satisfy those requirements.