Seedance 2.5 Pricing

Seedance 2.5 pricing is based on input type, resolution (480p, 720p, or 1080p), aspect ratio, and video duration (4–30 seconds). You pay a per-second rate for the seconds of video the model processes.

There are two input modes, and they bill on different amounts of video:

• •Text / Image-to-video: generate from a text prompt, a first-frame image, and/or reference images. You are billed for the video you generate. Text prompts and still images do not add a meaningful amount.
• •Video-to-video: generate using one or more reference videos. You are billed for the video you generate plus the reference video the model reads. Every reference video counts, and their lengths add up. See the video-to-video section below before you budget.

Text / Image-to-video: Per-Second Rates

Multiply by your clip duration to get the total cost. These rates are per second of generated output.

Video-to-video: Per-Second Rates

Important: video-to-video bills your reference videos as well as your output. The rate below is lower per second than text / image-to-video, but it applies to more seconds: your output plus your reference footage. Budget with:

billed seconds = output duration + max(total reference duration, ⅔ × output duration, rounded up)
cost ≈ rate × billed seconds

• •Total reference duration is the combined length of all reference videos you attach. Two references of 15s and 5s count as 20 seconds.
• •There is a minimum. The reference portion is never billed at less than two-thirds of your output duration, rounded up to the next whole second (e.g. 3s for a 4s output, 7s for a 10s output, 20s for a 30s output). A reference shorter than that is billed at the minimum.
• •There is no cap. Long references are billed at their full length.

This applies at every resolution, including 1080p. It is how the model bills, not a property of one tier.

Reference seconds are charged at the same per-second rate as your output (your output's resolution and aspect ratio), whatever the reference's own resolution or frame rate, so a lower-quality reference is not a cheaper one. A shorter one is, down to the minimum.

Is video-to-video cheaper than text / image-to-video? Usually not. With a reference as long as the output, video-to-video costs about 20% more than generating the same clip from text or an image. With a short reference (at or below the minimum), the two cost about the same.

Choose video-to-video for the motion and timing it transfers, not to save money.

Quick Cost Examples

Video-to-video rows show the total reference duration and the seconds actually billed.

Credit Holds vs. Your Final Charge

If you watch your balance during a generation, you may see a larger amount deducted than the clip ends up costing. That is expected:

1. •On request: Segmind places a temporary hold on credits, sized from this model's recent average cost. This is an estimate, not the price of your clip.
2. •On completion: the hold is settled at the actual cost of your request, and the difference is returned to your balance right away.

So the number to reconcile against the tables above is your balance after the request settles, not the momentary dip while it runs.

Failed generations are not billed. If a request returns an error, the hold is rolled back in full and you are charged nothing.

Important Notes

• •Audio is free: enabling audio generation does not affect pricing; there is no audio term in the rate matrix.
• •Video-to-video bills output + reference seconds: the single most common source of a surprising bill. A 4s clip driven by 20s of reference video is billed as 24 seconds, not 4.
• •Multiple reference videos add up: each extra reference adds its full length to the bill (above the minimum).
• •Output length scales linearly: doubling your output duration doubles that part of the cost, and raises the reference minimum with it.
• •Resolution is the main cost lever: 480p → 720p is roughly 2.2× per second, and 720p → 1080p is roughly 2.5×.
• •Aspect ratio is not a meaningful cost lever: every ratio at a given resolution prices within about 1% of the others. Choose the ratio your platform needs, not the cheapest one.
• •Adaptive aspect ratio follows your first frame and is billed at the resolved output ratio (a 16:9 first frame bills at the 16:9 rate).
• •Rates are close estimates: the final charge is computed from the exact amount of video processed and can differ from these figures by 1–2%.

Cost Optimization Tips

• •Trim your references to the part you need: reference seconds are billed, so cutting a 15s reference down to the 5s that matters saves 10 seconds of billing (down to the minimum of two-thirds of your output length).
• •Attach only the references you need: every extra reference video adds its length to the bill.
• •480p for drafts, 720p for delivery, 1080p for hero shots: iterate at 480p, render standard finals at 720p, and reserve full-HD 1080p for the clips that need it (~2.5× the 720p cost).
• •Match duration to the shot: cost is strictly per second of video, so trim to the length you need (4–30s).
• •Choose aspect ratio by platform: 16:9 for YouTube, 9:16 for Reels/TikTok/Shorts, 1:1 for social square. The cost difference is negligible.
• •Fix the seed to iterate: lock a seed and refine the prompt instead of re-rolling full-length renders.

Example Scenario

A team producing 30 clips/day at 9:16, 720p, 5 seconds (text / image-to-video) spends about $35.84/day ($1.1946 × 30), or roughly $1,075/month. Rendering the same clips at 1080p would cost about $88.14/day ($2.9378 × 30).

The same 30 clips/day as video-to-video with a 5-second reference each bill 10 seconds apiece, roughly $42.87/day at 720p ($1.429 × 30). With a 15-second reference each, they bill 20 seconds apiece, about $85.74/day. Trimming references to the part you need is the easiest saving.