Best for
- Product showcase and ad clips
- Text-to-video and image-to-video tests
- 15 or 30 second storyboards when the route supports them
Video API · Seedance · separate site
Seedance 2.5 Video API: Async Generation and Pricing Checks
Seedance 2.5 is a video workflow, not a chat completion. Create an async task, save the task ID, poll to a terminal state and reconcile the final charge after success.
When this model is a good fit
Use Seedance when a longer, more coherent video shot matters more than instant text response. Open the separate Seedance site for its own account and billing. This page is intentionally explicit about account separation and task-state billing.
Check before production
- Open the separate Seedance site before creating an account
- Save task ID and idempotency information
- Download only after a successful terminal state
What this page does not guarantee
Do not assume the main-site account, balance, API key or billing ledger can be used on sd.gpt345.com. They are separate.
How to choose this hub
Choose Seedance when the deliverable is a short video shot and the workflow can wait for an asynchronous task. The key decision is duration and task reliability, not whether a chat-style endpoint responds immediately.
| Option to compare | Choose it when | Measure before switching |
|---|---|---|
| Grok Imagine Video | You want to compare another creative video route on the main site. | Prompt adherence, image-to-video behavior and completed-task cost |
| Seedance 2.0 or a fast tier | You are prototyping many short ideas before a final render. | Time to first usable clip, failure rate and duration rules |
| A manual editor | The shot requires frame-level control or brand-safe revision. | Human editing time versus the number of accepted generations |
Production workflow
- Confirm you are using the separate Seedance site and its own account before creating a key.
- Create one short task with a unique client reference and save the task ID.
- Poll the task instead of posting a duplicate request after a client timeout.
- Download after terminal success, then reconcile duration and charge in the Seedance console.
Pricing and billing
The current campaign may show Seedance 2.5 reference pricing such as 1 second / 0.1 RMB. Treat campaigns as time-limited and verify the live Seedance console before production.
Failed and cancelled tasks should be checked against the task ledger. A timeout is not proof of failure; query the task before submitting again.
Minimal API request
Use the shared gateway and keep the model ID in configuration. Start with a small request, save the request ID, and compare usage with the dashboard before increasing concurrency.
curl https://www.gpt345.com/v1/videos/generations -H "Authorization: Bearer $GPT345_API_KEY" -H "Content-Type: application/json" -d '{"model":"seedance2.5","content":[{"type":"text","text":"A clean product reveal with a slow camera move"}],"duration":15}'Acceptance checklist
- The task returns an ID and an observable status.
- The terminal status is success before download.
- The duration and successful-task charge match the Seedance console.
Frequently asked questions
Do main-site and Seedance accounts share data?
No. Accounts, balances, API keys and billing are separate.
Should I retry after a client timeout?
Check the task ID and status first. Blindly posting again can create a second billable task.
Is the 1 second / 0.1 RMB campaign permanent?
No. Treat it as a campaign reference and confirm the current offer in the live console.