Seedance 2.0 Actual Billing: Output Tokens First, Per-Second Cost Afterward#

Customers often ask a simple question: how much does Seedance 2.0 cost per second? The accurate answer is that Seedance 2.0 is not billed by a universal fixed second rate. It is billed by actual output tokens returned after the task completes.

Core takeaway#

The correct billing flow is:

It is not:

So we should not say:

Instead, after the task finishes, we can calculate:

This is an observed value for that task, not a universal fixed rate.

Current public capability boundary#

Current public capability boundary for doubao-seedance-2-0 and doubao-seedance-2-0-fast:

• 480p supported
• 720p supported
• 1080p is not currently in the public supported range

The measured examples below use 720p and 4s.

Billing rules#

Seedance 2.0 billing mainly depends on whether the request contains video input.

video0 means there is no video reference input. video1 means the request includes video reference input, such as reference_video.

Final billing formula#

After a successful task, the system first reads TotalTokens. If unavailable, it falls back to CompletionTokens.

Without extra multipliers or discounts:

Crazyrouter quota conversion:

Measured case 1: text to video#

Request profile:

• Model: doubao-seedance-2-0
• Type: text-to-video
• Input: text only, no image or video reference
• Resolution: 720p
• Duration: 4s
• Task ID: cgt-20260617212928-5s755
• completion_tokens = 87300
• total_tokens = 87300

Because there is no video reference input, the billing key is:

Calculation:

Result: this simple 4-second 720p text-to-video task was observed at about $0.14/sec.

Measured case 2: reference video generation#

Request profile:

• Model: doubao-seedance-2-0
• Type: reference-video generation
• Input: text + previous generated video as reference_video
• Resolution: 720p
• Duration: 4s
• Task ID: cgt-20260617214300-rsnsx
• completion_tokens = 173700
• total_tokens = 173700

Because the request contains video reference input, the billing key is:

Calculation:

Result: this 4-second 720p reference-video task was observed at about $0.17/sec.

Comparison of the two tests#

This comparison shows why a single per-second price is misleading: video-input requests have a lower unit token price, but may return more tokens, so the final observed per-second cost can still be higher.

Customer-facing explanation#

Seedance 2.0 does not have a fixed per-second price. It is billed by actual output tokens after the task completes. We can calculate an observed average cost per second from completed tasks, but there is no universal fixed tokens-per-second conversion.

If customers ask why there is no fixed second price, the reason is simple: duration is a task parameter, but the final billing unit is the actual output tokens returned after generation.

Internal recommendation#

To give customers more stable estimates, collect more real task samples and group them by:

• text-to-video without reference input
• image-to-video
• reference-video generation
• with or without audio
• duration
• prompt complexity
• resolution

Then calculate p50, p75, and p90 tokens/sec. This is much more reliable than guessing a universal tokens/sec number.

Bottom line#

The safest customer-facing explanation is: Seedance 2.0 is billed by actual output tokens. In the current 720p/4s tests, simple text-to-video was about 0.14/s,whilereference−[videogeneration](/en/topics/video−generation)wasabout0.14/s, while reference-[video generation](/en/topics/video-generation) was about 0.17/s.

For one API layer across video, image, chat, and other AI models, try Crazyrouter.