Seedance 2.5 token formula in Python: price any clip before you send

Seedance 2.5 does not charge per second. It charges per video token, so a 10-second clip costs a different amount at 720p than at 1080p, and a 30-second clip is not a flat multiple you can guess. The fal listing for the model states the rule: $0.0214 per 1,000 tokens, where tokens = (output height x output width x duration in seconds x 24) / 1024 (read 2026-10-05). Sume bills the provider list times 1.25, rounded up to the cent, so you can quote a clip before you submit it.

The formula, with Sume's frame sizes

Sume's pricing code uses fixed frame sizes per resolution and aspect ratio, not your exact pixels. At 720p and 16:9 that is 1280 x 720, which gives 21,600 tokens per second of output (1280 x 720 x 24 / 1024 = 21,600). Multiply by seconds, then by $0.0214 per 1,000 tokens, then by 1.25.

A function you can run

This quote helper covers 720p and 16:9 only, because that is the one frame size shown above. It prints the same numbers as the table. Extend the size dictionary from the catalog when you need other ratios.

import math

RATE_PER_1000 = 0.0214  # Seedance 2.5 provider list, 480p/720p
MARGIN = 1.25
SIZES = {("720p", "16:9"): (1280, 720)}


def quote(resolution, ratio, seconds):
    width, height = SIZES[(resolution, ratio)]
    tokens = math.ceil(width * height * seconds * 24 / 1024)
    provider = tokens / 1000 * RATE_PER_1000
    billable = math.ceil(provider * MARGIN * 100) / 100
    return tokens, round(provider, 2), billable


for seconds in (5, 10, 15, 30):
    print(seconds, quote("720p", "16:9", seconds))

Where it can drift

Treat the output as a quote, not a ledger. The real reserve happens at submit, and the poll response reports the amount Sume actually billed in usage.cost. The catalog at GET /v1/video-router/models is the source of truth for rates; the 1080p rate for seedance-2.5 ($0.0234 per 1,000 tokens in Sume's pricing code) is higher than the 720p one.