Credits and pricing
How credits work across Seedance, Kling, image models, and video routes
Understand shared credits, subscription credits, credit packs, higher-cost model routes, and failed-generation credit returns.
Credits are the shared balance for video and image generation. The estimate should be shown before submission and can change by model, quality, duration, audio, reference count, and provider route.
Quick decision
Best for
- Users comparing plans, credit packs, high-cost model routes, and failed-task credit returns before paying.
- Teams that need predictable credit planning across video, image, reference, and upscale work.
Not ideal for
- Users who expect every model, resolution, duration, and reference count to cost the same.
- Workflows that cannot tolerate provider-cost-based estimates.
Choose this when
- You need to understand why a Seedance draft, Kling route, 1080p video, audio task, or image model can use different credits.
One shared balance
Subscriptions and credit packs use the same credit balance. A creator can use credits for Seedance-style video, Kling routes, image models, upscales, references, and overflow work.
Why higher routes cost more
1080p video, audio-enabled video, image-conditioned video, and expensive third-party provider routes can cost more than fast draft generation. The workspace should make that visible before the job starts.
Safe credit consumption
A failed generation should not silently consume the user's balance. The intended behavior is automatic credit return when a task fails before a usable output is delivered.
Quick answers
How are credits calculated?+
Credits are estimated from provider cost, selected model, duration, quality, audio, reference count, and route type. The estimate should appear before generation.
Do higher plans make credits cheaper?+
Higher-value subscriptions and credit packs can include bonus credits, so the effective cost per credit is lower than the entry plan.
What happens if generation fails?+
Failed jobs are designed to return credits automatically, so users do not pay for a task that never produces a usable result.