A long wait and a failed generation are different situations. Start by checking the request status: it may still be queued or processing.
If generation is still running
- Allow time for the status to update. Waiting time depends on the model, duration, quality and service load.
- If the page lost its connection, reopen the conversation through History or refresh the tool page.
- Check whether a result has appeared. Do not submit a duplicate solely because the page has not updated.
If an error appears
- Open the error details and follow the suggested action if one is provided.
- Check whether the model supports your files and settings. Replace an unsuitable file or adjust settings.
- Check your subscription, available balance and workspace limits.
- If a request was rejected because of its content, revise it according to the stated reason.
- For a temporary service failure, try later. Retry only after the previous request has finished and check the new run’s price.
What happens to tokens?
Find the operation in usage history. An active reservation, a final charge and a released reservation represent different states. A temporary reduction in available balance does not necessarily mean tokens are permanently lost, and refunds are not identical for every error.
Details to keep for support
- The start time and tool or model name.
- A conversation link or generation identifier, if available.
- A screenshot of the error and a description of your steps.
- The usage history operation if your question concerns the balance.