GPT Image 2.5 is live — OpenAI's newest image model, targeted edits that leave the rest of the frame alone
All resources

· reAPI Team

fal User is locked: check the account, key and balance

Diagnose fal User is locked errors by checking the response, account, credits and key permissions. Prepare support evidence before retrying or paying more.

fal User is locked: check the account, key and balance

User is locked can keep appearing after a top-up. In fal's public issue 922, users described that mismatch: the dashboard showed credit, while requests still received a lock response. The thread contains different reported fixes and failures; its maintainer ultimately directed affected users to account-specific support.[1]

That history does not establish the cause of your error. It does explain why “add more money” or “create another key” is an incomplete answer. Start with the full reason returned by the API, the account behind the running application and the time the error occurred. Pause the application's automatic submissions while those facts are unresolved.

Keep the word after “Reason”

An exhausted-balance message and an administrative-lock message should not be collapsed into the same log entry. Preserve the complete response body with its HTTP status, endpoint, method and timestamp. A log that retains only 403 has discarded information you will need for the next decision.

Then identify which operation failed. Was the application submitting a new generation, checking a request already accepted, or trying to retrieve its result? An SDK exception displayed at the end of a workflow may come from any of those steps. The remedy depends on the step that actually received the rejection.

Consider a hypothetical sequence: submission returns a request ID, the worker saves it, and a later poll receives the lock error. That sequence does not prove the earlier generation failed or was canceled. Keep the ID and the responses for both operations. Replaying the whole workflow would create a new submission without resolving the existing request's outcome.

By contrast, if the initial submission was rejected and no request ID arrived, record that specific result. Do not invent a task identifier or assign a final generation state from the authorization message. You are investigating whether the operation was allowed, not yet assessing the quality of generated media.

The fal completed-status guide covers the separate case where a known request reaches a terminal state but its media is missing. Keeping these cases distinct prevents an account problem from turning into an unnecessary regeneration.

Check the account before interpreting the balance

fal's FAQ says accounts lock when their balance falls below the account's lock threshold. It recommends adding credit for that case and treats enterprise invoice billing separately.[2] The rule is not a universal statement that any displayed positive amount guarantees access for every account.

First match the account that owns the application's key to the account selected in the dashboard. Include the team context where relevant. The browser can be showing a funded team while a worker uses a personal key; that is a troubleshooting possibility to check, not an explanation we have established for your request.

Inspect the running process's configuration through your normal secret-management tooling. A key name or safe identifier can help confirm the source without printing the secret. If someone recently changed the credential, verify that the deployment actually loaded the intended configuration. Updating a local environment file does not establish what a separate worker is using.

Compare the balance observation and error at known times. For example, a screenshot taken before the top-up cannot show whether the balance had changed when the later request failed. Record the receipt time, the dashboard capture time and the failed operation's time together. That gives support a sequence it can inspect instead of two undated screenshots.

If the affected account's balance and returned reason fit the documented funding threshold, follow that billing guidance. If a completed purchase is missing, investigate the purchase and credit record. If the correct account shows credit but the lock remains unexplained, another deposit is not evidence that the problem will resolve.

A dashboard login is also a different operation from API access. If you cannot sign in with GitHub because the one-time code never arrives, fal's FAQ says to check the GitHub account's primary email.[2] That can help you reach the account page; it does not explain why an already configured application receives an account-lock response.

A funded account can still have the wrong key policy

fal documents permission presets, custom permission lists, allowed-model restrictions and key expiration. Its API preset is intended for consuming hosted models.[3] Those controls answer what the credential may do. A balance answers what funding is present. Correcting one does not establish the other.

Compare the exact failed operation with the key's policy. If the key is limited to certain models, check the intended endpoint against that restriction. If it has expired, correct the credential rather than adding credit. If a client is sending a different key than expected, fix that configuration before changing the account's permissions.

A detail that can be easy to miss is how the policy was created. fal says a preset key follows the current preset definition, while a custom policy keeps its permission list from creation. Adding a model restriction stores the key as a custom policy.[3] Two keys created for the same application can therefore deserve separate inspection rather than an assumption that their permissions remain identical.

Do not give every key full access just to see whether the error changes. Grant the operations the application needs and preserve the previous policy in your investigation notes. A successful request after several simultaneous changes would leave you unsure which change mattered—and might leave broader access than the application requires.

Likewise, replacing a key is not a documented universal account-unlock operation. In issue 922, some users reported success after a new key or SDK update, while others said those steps did not work.[1] These are individual observations. They do not establish that an SDK version, key migration or delayed unlock is the cause of every similar message.

When comparing two clients, first make the intended account, model and operation match. A successful request for one model from a laptop does not establish access to another model from a scheduled worker. Record the remaining difference before choosing a correction. Otherwise, a change in model or account can be mistaken for a fix to the client library.

Review the installed client version and request construction if the evidence points there. Keep the full service response before changing packages. That makes it possible to distinguish a client configuration correction from an account restriction that needs the provider's attention.

Hosted models and Serverless deployment need different checks

If you only call fal's hosted Model APIs, the installation guide says you need an API key and the client SDK, not the Serverless CLI. Deploying your own Serverless apps requires account approval; calling hosted Model APIs does not require that approval.[4]

This matters when a search result sends you through CLI installation or team-switching steps. Confirm that those steps apply to the product and process receiving the error. A working CLI profile does not, by itself, establish which credential your separately deployed web service uses.

For browser applications, fal recommends keeping the key behind a server-side proxy.[2] Putting the secret into browser code is not a way to repair an account lock. Check the proxy's actual credential and returned error, keeping authentication on the server.

Give support one account and one sequence to investigate

For an unexplained lock, the maintainer's instruction is to contact fal support using the same email as the affected account. Lock cases are handled individually; a public issue comment cannot perform the unlock.[1]

A concise private report can follow this shape:

Account: [affected account]. At [UTC time], [operation and endpoint] returned [full sanitized error]. The application uses [safe credential identifier] belonging to [account or team]. At [capture time], that account displayed [balance]. Purchase [reference], if relevant, was recorded at [time]. Request [ID], if one exists, was accepted before the later error. Please review the account restriction and the operation shown here.

Add the client name and version, and say which checks you have already completed. Remove the authorization header, API key and unnecessary payment details. Describe observations plainly: the key used, the response received and the balance displayed. Avoid assigning a database fault or billing motive that those records do not establish.

Read the latest maintainer response in any issue you use as background. A separate report, issue 1163, was closed after a maintainer said support had resolved that account earlier.[5] That does not prove every account is fixed, just as the original report does not prove an outage is still happening now. There is no universal waiting time or refund promise to infer from either thread.

If support says the account is unlocked, return to the operation that failed. For the hypothetical accepted job followed by a rejected poll, check that existing request first. Restored access may let you retrieve its state; it does not guarantee that the generation itself succeeded. Save the new response beside the earlier lock message. If the same operation still fails, send that updated timestamp and response in the existing support conversation so the review concerns the current account state.

When access returns, review the paused work before resuming it. Separate submissions that were rejected from jobs that already received IDs, and inspect the latter before recreating them. A generation used to test restored access can incur a charge if accepted; this article has not run such a test. Resume only the work you still intend to perform, with the corrected account and credential recorded.

Keep these checks if you move the workload

If the lock leads you to evaluate another provider, look for the same separation between credentials and funding. On reAPI, an invalid key returns HTTP 401 with code 10003, a revoked key returns 10004, and insufficient credit returns HTTP 402 with code 30001.[6] A worker can read the balance of the account behind its key with GET /api/v1/balance without starting a generation.[7]

Clear codes do not replace the habits above. Keep the full response, the time and the operation that failed, and confirm which account the deployed key belongs to before adding credit.

References

  1. fal-ai/fal. Issue 922: Account locked despite positive balance, complete thread including maintainer guidance. Retrieved November 15, 2026.
  2. fal. Model APIs FAQ. Retrieved November 15, 2026.
  3. fal. Key-Based Authentication: permissions, allowed models and expiration. Retrieved November 15, 2026.
  4. fal. Installation & Setup: Model APIs and Serverless access. Retrieved November 15, 2026.
  5. fal-ai/fal. Issue 1163: Account re-locks with Exhausted balance despite positive balance, including maintainer resolution. Retrieved November 15, 2026.
  6. reAPI. Errors. Reviewed November 17, 2026. Error codes.
  7. reAPI. Balance. Reviewed November 17, 2026. Balance reference.
All resources