Billing and Statements

LocoPal's balance, plans, consumption, and statements are all centralized in the LocoPal Ai account. Creation is not billed separately; its consumption is counted toward the same account.

Billing Methods

Different models are billed in different ways; the details page in the model marketplace is authoritative.

Video models (such as doubao-seedance and the seedance series) show a separate price matrix, split by "whether video input is included" and "output resolution". When reading it, first determine whether this request submits video input, then choose the resolution you need:

• Uploading only text or image references → look at the "without video input" column;
• Submitting video as input → look at the "with video input" column;
• Unsupported combinations are explicitly marked "unsupported"; do not infer a tier that is not shown from adjacent tiers.

Settlement Timing

This is the point most likely to cause misunderstanding; please distinguish synchronous APIs from async tasks.

Key behaviors of async tasks:

• Before submission, the platform checks whether the account balance (or the current key's quota) can cover the estimated consumption for this request; if not, it returns a readable message directly and does not create the task;
• No intermediate charges are incurred while the task is running;
• Querying task status does not incur charges; only creating a task is billed;
• Billing for failed tasks is handled according to that model's and the platform's price configuration.

Balance and Checks

• Before a call, the platform checks whether the balance covers the estimated consumption; if not, it returns a message instead of creating a task;
• Estimates for text models are based on input length; the actual amount is determined by usage in the response;
• Image, video, and other per-call or per-duration billing usually have estimates consistent with the actual amount within the range the model supports;
• We recommend that your business backend implement pre-checks and friendly messages for insufficient balance, rather than relying entirely on errors returned by the platform.

Where to Verify

The wallet provides top-ups, plans, and benefit redemption; the model marketplace provides estimated prices; statements provide final settlement. The three serve different purposes: a displayed price does not equal the final charge.

Verifying an Unexpected Charge

Locate it in the following order; do not judge by feeling:

1. Determine the time range: first narrow it down to the period when the problem occurred;
2. Find the task or request identifier: use the task ID for async tasks and the request ID for synchronous calls;
3. Confirm the model: use the model ID (not the display name) to confirm which model produced the charge;
4. Compare usage and unit price: for text, compare the input and output tokens in usage; for video, compare the resolution, duration, and input type;
5. Check for duplicate submissions: the most common anomaly with async tasks is that a task was created one extra time, rather than a unit price error.

After completing these steps, contact platform support and provide this information. Do not provide the full key.

Controlling Consumption

• Split keys by application: create separate keys for the production service, test environment, and customer applications so you can trace the source of consumption;
• Set quotas and validity periods for keys: limit the impact of abnormal consumption;
• Choose the right model: a smaller model often significantly reduces cost for the same task;
• Avoid duplicate submissions: save the task ID for async tasks and track progress by querying rather than recreating;
• Transfer results promptly: return URLs for images and videos may be short-lived, and you need to regenerate them after they expire.

Common Misconceptions