FAQ Summary

This page summarizes frequently asked questions about using the platform by topic. Each entry gives the cause and an actionable next step; for more detailed API-level troubleshooting, see Error Codes and Troubleshooting; for workspace issues, see Web Creation FAQ.

Account and Sign-in

Why does opening Creation ask me to sign in again?

LocoPal Ai is the single source of identity. Creation relies on the cross-subdomain session cookie issued by locopalai.com; if that cookie is invalid or missing, the sub-platform redirects back to the unified sign-in page.

First check: is LocoPal Ai signed in? If it is signed in but the sub-platform still shows you as signed out, refresh the sub-platform page; if that still does not work, sign out of LocoPal Ai and sign in again.

Do I need to sign in to Creation separately on a new computer?

No. Sign in once to LocoPal Ai, and Creation automatically receives the signed-in state.

Does changing my password or signing out affect other platforms?

Yes. Signing out, switching accounts, and changing your password all invalidate Creation in sync; this is normal behavior for a shared session.

API Keys

Can one key call all APIs?

Yes. The model, image, video, audio, and asset APIs share the same key; there is no need to create a separate key for each API.

Why can't this key see a certain model?

The visible scope of models is determined by the call group the key belongs to. Use that key to request GET /v1/models and confirm: if the model is not in the returned data, it is not within the authorization scope of this group.

What does "data": [] mean?

It means there are no available models under the group the current key belongs to. First confirm that the group configuration is correct; if you still have questions, contact platform support and provide the key name (do not provide the full key).

What if a key is leaked?

Immediately disable the key in "API Keys" in the console (rather than waiting for it to expire naturally), then create a new key for your business and replace the configuration. Disabling does not affect the assets, tasks, or history under the account.

Why shouldn't a key be written into frontend code?

Anything in a browser or a mobile app package can be extracted. The key must be stored in server-side environment variables or a secrets management service, with your business backend calling the platform on its behalf.

Models and Calls

What should I put in the model field of a request?

Fill in data[].id returned by GET /v1/models, exactly as returned. Do not use the display name shown on the model marketplace page.

"Model temporarily unavailable" or "no available channel" is displayed

It means the platform cannot route to that model at the current time; it does not mean the key is at fault. First refresh the model list or switch to a model of the same kind; if it remains unavailable, contact platform support and provide the request ID.

A text call returns empty content and finish_reason is length

A reasoning model may consume all of max_tokens in the reasoning process (completion_tokens_details.reasoning_tokens). Raise the output limit, or switch to a non-reasoning model and retry.

The image API reports unknown_parameter

Some image models (such as the GPT Image series) do not accept the response_format field. Remove that field and retry. Support for size, quality, and response_format is subject to the current model.

The returned image URL won't open

The image API may return a short-lived URL; transfer it to your own storage as soon as possible. Images that need long-term reuse should first be ingested into the asset library, then referenced via asset://<asset_id>.

Video Tasks

Why is video asynchronous?

Video generation takes a long time, so the platform handles it with a "create task → query result" approach. The create API returns a task ID, and you then poll for progress and results with the query API.

If the local wait times out after submission, should I create the task again?

No. After you have the task ID, even if the local wait times out, you must continue querying the original task. Creating it again produces duplicate tasks and duplicate charges.

The create response has no progress; did it fail?

No. The create response only needs to rely on id / task_id, status, model, and created_at; fields such as object and progress may not appear until the first query, so they must not be treated as requirements for successful creation.

The status is failed; how do I see the reason?

Read error.code and error.message in the same response. A terminal video failure is usually video_generation_failed; error.message is already a cleaned-up, readable reason, so do not parse the JSON inside it again.

Does querying bill repeatedly?

No. Queries only read status and incur no charge. Only creating a new task produces a new charge, which is why you should not resubmit repeatedly.

Assets and Portrait Authorization

Do I need to upload assets for each model separately?

No. Assets only need to be ingested once, into the platform's unified asset library; every model that supports asset references uses the same asset, the asset ID stays the same, and switching models only requires changing the model field in the request.

When can an asset be used?

An asset is reliably usable only after data.asset.status becomes Active. When it shows "Uploading" or "Processing", keep it and wait; do not repeatedly submit the same failed item in quick succession.

What image URL should I use to submit an asset?

Use an http(s) URL that can be downloaded directly from the public internet, or submit data_url according to the API's restrictions. Do not use a browser's local file path as an API parameter.

Can ordinary images replace portrait assets?

No. A portrait asset must be liveness-captured and authorized by the person themselves by scanning a QR code with a phone. The console flow is "Asset Library Management → New Group → Portrait Assets → Generate QR Code → The person scans it → Scan complete, start sync".

What happens when I switch to a model that does not support portrait assets?

Portrait assets only take effect on models that explicitly support them. When you switch to an unsupported model, the platform does not automatically downgrade them to ordinary reference images; use virtual assets instead or switch back to a supported model.

Do assets affect other models?

Ingestion is independent of models. Once an asset's status becomes available, it can be used with any model that supports asset references; there is no need to wait for preparation on the model side. Existing assets are prepared once by the platform when first referenced.

Billing and Balance

Does submitting a video task charge me?

No. Async tasks are settled on completion; a successful submission only accepts the task and does not pre-charge. Before submission, the platform checks whether the balance or key quota can cover the estimate for this request; if not, it returns a readable message directly and does not create the task.

Does a parameter validation failure (400) incur a charge?

Usually not. 400 means the request was not accepted and generally does not incur a generation task charge.

Is the price on the price card the final charge?

Not necessarily. Price cards are for estimation; the actual charge is determined by the request's actual usage, its group, and the server-side billing records. See Billing and Statements for details.

I think a charge is wrong; how do I verify it?

Use the task ID, request identifier, model ID, and time range to locate the specific record, then compare it against the statement details. Do not provide the full key.

Platforms and Creation Tools

What is the relationship between the two platforms?

LocoPal Ai is the core, providing accounts, billing, models, assets, and task records; Creation is a client that connects to it. See Platform Overview for details.

Can assets generated in Creation be used via the API?

Yes. Both share the same asset library, and asset IDs are the same.

Does refreshing the page regenerate or charge again?

No. Refreshing only restores the page state; it does not recreate tasks or bill again. Do not resubmit tasks that are still running.

Still Not Resolved

When submitting a support request, please provide: the time it occurred, the model name, the task ID or request ID, the HTTP status (for API calls), the sanitized error message, and the request type (text / image / video / audio).

Do not submit: API keys, the full Authorization request header, sign-in cookies, real portrait materials, or full signed asset URLs. For the relevant scope, see Error Codes and Troubleshooting.