Account, credits, and troubleshooting

Troubleshooting and support

Check upload, generation, image display, access, and payment problems by symptom.

On this page

Start with the symptom

Symptom	What to check
Generate is disabled	Check unfinished uploads, prompt length, reference count, permissions, or an active submission. Resolve the displayed reason instead of clicking repeatedly.
Task stays queued	Queued is not failed. Wait for existing tasks and stop submitting more when too many are pending.
Content did not pass moderation	Adjust prompt and references. Do not repeatedly resubmit unchanged content; a content block and moderation-service failure have different charging treatment.
Image failed to load	Use the image retry control and check connectivity. Successful generation does not mean the device has downloaded its preview.
Missing model or option	Check mode, capabilities, and membership. Models offer different sizes, quality, and tools.
Share unavailable or read-only	Check account identity, expiry, disabled status, and changes to permissions.
Payment made but benefits unchanged	Check order status and use its update control. Keep the order number for support; do not pay again immediately.
Blank page or mobile layout issue	Check connectivity and refresh. Report device, browser, and steps with a full-page screenshot. Include simulated viewport and zoom when using device simulation.

Too frequent does not necessarily mean you clicked too fast

An image-generation or AI-chat rate-limit notice means the model service rejected this request. Models and routes have different limits on requests, image counts, usage, or simultaneous work, and other users may share the same service allowance.

Currently these generation API rate limits appear as failures, without an exact waiting countdown or automatic switch to another model. Stop repeated submissions and try later. For recurring failures, report the model, time, and full message.

The site's queue allowance, running task count, and provider API rate limits are different conditions. Queued work is still in progress. An explicitly failed image request releases relevant reserved credits; failed chat releases the relevant reservation. Resubmission is a new request.

Generation failure, result retrieval, and preview loading differ

A progress bar alone cannot distinguish model generation from image retrieval or establish whether the provider has charged. Check the original task's final status and give Support the complete message when investigation is needed.

What you see	What it means	Next step
Task is queued or generating	There is no final status yet. No image yet does not mean failure.	Wait and check the original task instead of submitting more requests for the same operation.
The model returned a result but image retrieval is temporarily failing	The system can retrieve the same result without calling generation again for a download problem. Recovery is not unlimited or guaranteed.	Wait while the task remains in progress. If it fails, record the full message and report it before deciding to regenerate.
Generation succeeded but a thumbnail or preview cannot load	Task success differs from this device loading its image. The original may already be saved while preview requests fail because of connectivity.	Retry the image or refresh to check status, then check connectivity. Do not immediately regenerate.
Task explicitly failed	Check the cause. Content blocks, service errors, insufficient credits, and invalid parameters differ.	Resolve the cause and check credit records before submitting a new request.

A partial reply or a disconnected page

Ordinary model or upstream network failures may leave approved, saved partial text and a failure notice. Partial text is not a complete answer. Explicit moderation blocks clear the turn's reply and thinking; refreshing does not restore them.

Closing the page or losing local connectivity does not cancel an accepted generation. Reconnect, return to the original conversation, and refresh to check saved status and content before sending again.

A platform restart, interrupted connection, and completed generation differ. Not every interruption can resume from the same point; use the original record's final status.

Before retrying

Record the error first, then decide whether to wait, adjust settings, or submit again. Regeneration and rerunning paid tools create new requests and may consume credits again.

Refreshing does not fix insufficient credits, storage, or unsupported parameters. Resolve those conditions first. Refresh to confirm a task's saved status when synchronization fails; an interrupted request is not necessarily failed generation.

Report a problem that can be located

Support email: support@bananai.com.cn.

Registered users can submit Feedback from the personal area and view replies or add details in Settings. Guests can use the support email in the page footer.
Include time, device and browser, feature, steps, and the complete error. Include an order number for payment issues or an account ID for account issues.
Keep only necessary screenshot information. Hide passwords, codes, contact details, payment information, and private images. Never send browser-session information.