Seedance 2.5 with a reference video: 10 s at 720p costs $8.67

A Seedance 2.5 request that includes a reference video is priced on more seconds than you asked for. On Sume, a 10-second 16:9 720p clip is $5.78 with no reference video and $8.67 with one. Reference images and reference audio do not trigger this in the pricing code; only reference video URLs do.

The rule is short. fal's Seedance 2.5 explainer, read 2026-10-08, says reference-video generations receive a 0.6x multiplier on the token bill. Sume's estimator assumes 15 seconds of reference video input, adds it to the output seconds, prices the total tokens, applies 0.6, then adds the 1.25 house margin and rounds up to the cent.

The same clip with and without a reference video

Tokens are width x height x seconds x 24 / 1024. The 720p, 16:9 frame is 1280 x 720, so 21,600 tokens per output second, at a list rate of $0.0214 per 1,000 tokens.

Where the crossover sits

The multiplier is 0.6 x (seconds + 15) / seconds. It equals 1.0 at 22.5 seconds, so a request of 22 seconds or less costs more with a reference video, and 23 seconds or more costs less. At 30 seconds the reference-video request is 0.9x the plain one. That looks like a discount, but it is an artefact of a fixed 15-second assumption, so do not add a reference video only to save money.

• At 480p a 10-second 16:9 clip is $2.69 plain and $4.04 with a reference video.
• At 1080p a 10-second 16:9 clip is $14.22 plain and $21.33 with a reference video.
• 1080p uses a higher list rate ($0.0234 per 1,000 tokens) than 480p and 720p ($0.0214), which is already in those figures.

What to do with it

If the reference is a motion or camera clip you need, pay the surcharge and draft at 480p first. If you only need a look, send reference images instead: they carry style and subject without the reference-video bill. The Sume request takes references through input_references entries of type image_url, video_url or audio_url on /v1/videos.

The fixed 15-second assumption is a reserve and pricing rule in Sume's code, not a measurement of your file. fal lists Seedance 2.5 reference limits of 30 images, 10 videos and 10 audio files with the video and audio pools each about 30 seconds in total. I found no Seedance-specific count limit in Sume's own request validation, so those provider limits are the ones to plan for.

Submit it and read the result

Sume reserves the billable amount from your workspace USD balance when the job is submitted, then returns a polling_url. Send one plain 480p job and one with a reference video, then compare the two charges to the figures in the table before you queue a batch. Poll the job until status is completed or failed; a completed job lists its files in unsigned_urls.

import os
import time
import requests

H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
body = {
    "model": "seedance-2.5",
    "prompt": "A slow push-in on the scene, soft natural light",
    "duration": 10,
    "resolution": "720p",
    "aspect_ratio": "16:9",
}
r = requests.post("https://api.sume.com/v1/videos", headers=H, json=body, timeout=60)
r.raise_for_status()
url = r.json()["polling_url"]
while True:
    job = requests.get(url, headers=H, timeout=30).json()
    if job["status"] in ("completed", "failed"):
        break
    time.sleep(10)
print(job["status"], job.get("unsigned_urls"))

Check the quote before a batch

Sume reserves the billable amount from your workspace USD balance when you submit. Run one clip at 480p, compare the charge with the figures above, and then multiply for the batch.