The cost of a Seedance 2.5 API workflow depends on what the application needs to create and deliver. A native model request, a standalone enhancement job and a combined Production workflow use different pricing logic.
Before comparing prices, define four things: the model workflow, target duration, final resolution and whether the request contains video input. This avoids misleading “price per video” comparisons and produces a more useful production estimate.
Three Seedance-related workflows
Native Seedance 2.5 API
Native Seedance 2.5 is the right option when a team wants direct access to the model contract. It supports native 480p, 720p and 1080p video, multimodal references, native audio, editing and extension for supported request modes.
Native generation follows the applicable model and usage pricing. The final cost can vary with model mode, delivered output and the commercial terms assigned to the customer.
Standalone Video Enhancement
Standalone enhancement processes a video that already exists. Pricing is based on the selected enhancement tier, output resolution, actual enhanced duration and output FPS band.
The output FPS multipliers are:
| Actual output FPS | Price multiplier |
|---|---|
| Up to 30 FPS | ×1 |
| More than 30, up to 60 FPS | ×2 |
| More than 60, up to 120 FPS | ×4 |
The final enhanced duration is rounded up to a whole second. If FPS is omitted, the request reserves at the highest FPS band and then settles using the actual output FPS.
LSF Seedance2.5 Production
Production combines Seedance 2.5 creative control and AI finishing for 1080p, 2K or 4K delivery in one task. It uses three tiers—Lite, Standard and Pro—and two input-based rate types.
Production currently has no additional FPS price multiplier.
Production pricing without video input
For text, image, frame or audio-only requests, billing is based on the final enhanced output duration rounded up to whole seconds.
| Tier | 1080p per billable second | 2K per billable second | 4K per billable second |
|---|---|---|---|
| Lite | $0.79 | $0.82 | $0.89 |
| Standard | $0.82 | $0.89 | $0.99 |
| Pro | $0.99 | $1.25 | $1.69 |
Formula:
final charge = selected rate × ceil(final enhanced output seconds)
For example, Standard 1080p costs $4.10 when the completed output contains exactly five billable seconds. If the actual result is 5.042 seconds, it is billed as six seconds: $0.82 × 6 = $4.92.
Production pricing with video input
When a request contains one or more reference videos, editing input or extension input, Light Speed Future measures the input-video duration and adds it to the rounded output duration.
| Tier | 1080p per combined second | 2K per combined second | 4K per combined second |
|---|---|---|---|
| Lite | $0.4740 | $0.4920 | $0.5340 |
| Standard | $0.4920 | $0.5340 | $0.5940 |
| Pro | $0.7425 | $0.9375 | $1.2675 |
Formula:
final charge = selected video-input rate × (measured input-video seconds + ceil(final enhanced output seconds))
Images and audio references do not add input-video seconds. Multiple video references are measured and added together without rounding each input separately.
For example, a Standard 1080p request with a 5.000-second video reference and a 5.042-second final output uses 11 combined billable seconds: $0.492 × (5 + 6) = $5.412.
Why the reservation and final charge can differ
AI video jobs run asynchronously. Light Speed Future reserves an estimate when the request is accepted, then reconciles the amount when the completed output profile is known.
The final output can be slightly longer than the requested duration. Automatic-duration requests can also produce a conservative initial reservation. The reservation is therefore an estimate, not a spending cap.
Store the task ID and client_request_id, continue polling the original task and use the terminal result for reconciliation. Repeated GET requests do not generate the video again or duplicate the charge.
How to estimate a real production workload
Use a benchmark set that represents your application rather than relying on one showcase prompt.
- Separate text/image requests from requests containing video.
- Test the durations your users are likely to submit.
- Compare Lite, Standard and Pro only where those tiers are enabled.
- Record the final duration, resolution and FPS returned by completed jobs.
- Track usable-output rate as well as raw completion rate.
- Estimate monthly spend using the final reconciled usage from the benchmark.
This produces a more realistic forecast for social content, e-commerce, agency production or customer-facing AI video products.
Choose the right pricing model
Choose native Seedance 2.5 when direct native generation is the priority. Choose standalone enhancement when you already have a source video. Choose LSF Seedance2.5 Production when your application needs one request to generate and deliver a finished high-resolution MP4.
Review current LSF pricing or contact the team for enabled product combinations and customer-specific commercial terms.