ByteDance sells Seedance by the token, and almost everyone else resells it by the second. The official conversion is simple once written down: tokens = (input seconds + output seconds) × width × height × 24 ÷ 1024, which at 16:9 is about 9,607 tokens per second at 480p, 21,600 at 720p and 48,600 at 1080p. Every per-second price on this page is that formula applied to each provider’s rate, checked on 2026-09-30, for jobs with no reference video.
The short version, stated plainly because most comparison pages do not: for Seedance 2.5 and 2.0 the most affordable place is BytePlus ModelArk itself, and OpenRouter and Replicate pass that list price through almost unchanged. APIMODELS is about 17% above list on 2.5 and 31–33% above on 2.0, and below fal, Kie.ai and EvoLink on both. The one tier where APIMODELS is under the official list is Dreamina Seedance 2.0 Mini, at $0.0225 per second at 480p and $0.048 at 720p.
Price is not the only axis. A BytePlus account means an enterprise sign-up and per-model activation, and the official Ark line rejects any reference image that shows a real person unless it has first been registered in the asset library. On APIMODELS the seedance-2.0 models accept a real-person photo passed as a plain https URL (a rejected upload is registered and retried automatically), one USD balance covers every model, and there is no account to open anywhere else.
Seedance 2.5 generates 4 to 30 seconds with synced audio. The official 1080p rate is $0.569 per second, but most resellers charge $0.74 to $1.16 for it; APIMODELS currently offers 2.5 at 480p and 720p only.
| Provider | 480p | 720p | 1080p |
|---|---|---|---|
| BytePlus ModelArk (list) | $0.103 | $0.231 | $0.569 |
| OpenRouter | $0.103 | $0.231 | — |
| Replicate | $0.103 | $0.231 | — |
| APIMODELS | $0.120 | $0.270 | — |
| EvoLink | $0.138 | $0.296 | $0.739 |
| Kie.ai | $0.140 | $0.315 | $0.790 |
| fal | $0.221 | $0.473 | $1.164 |
Seedance 2.0 runs 4 to 15 seconds. It is the tier to use for 1080p, since 2.5 is not offered at 1080p here.
| Provider | 480p | 720p | 1080p |
|---|---|---|---|
| OpenRouter | $0.067 | $0.151 | $0.374 |
| BytePlus ModelArk (list) | $0.070 | $0.150 | $0.370 |
| Replicate | $0.080 | $0.180 | $0.450 |
| APIMODELS | $0.092 | $0.197 | $0.492 |
| EvoLink | $0.093 | $0.199 | $0.497 |
| Kie.ai | $0.095 | $0.205 | $0.510 |
| fal | $0.141 | $0.303 | $0.682 |
Fast and Mini stop at 720p. Both have an official promotion until 7 October 2026 (Fast 25% off, Mini 60% off, enterprise accounts only and capped), which OpenRouter currently passes through for Fast; the table uses list prices. On APIMODELS there are two lines for each tier: the seedance-2.0 line accepts real-person references, the Dreamina line (the BytePlus international version of the same model) does not but costs less.
| Provider | Fast 480p | Fast 720p | Mini 480p | Mini 720p |
|---|---|---|---|---|
| BytePlus ModelArk (list) | $0.060 | $0.120 | $0.040 | $0.080 |
| OpenRouter | $0.040 (promo) | $0.091 (promo) | $0.034 | $0.076 |
| Replicate | $0.070 | $0.150 | $0.040 | $0.090 |
| Kie.ai | $0.059 | $0.124 | $0.019 | $0.041 |
| fal | $0.113 | $0.242 | $0.072 | $0.155 |
| APIMODELS, seedance-2.0 line | $0.071 | $0.153 | $0.044 | $0.095 |
| APIMODELS, Dreamina line | $0.0644 | $0.137 | $0.0225 | $0.048 |
If you can open a BytePlus account and never use real faces, go direct for 2.5 and 2.0; nothing undercuts list on those tiers except promotions. For drafts and volume, Dreamina Seedance 2.0 Mini on APIMODELS at $0.0225 per second is a 15-second 480p clip for $0.34. For real-person work, use seedance-2.0 or seedance-2.0-mini here and pass the photo as a URL. For 30-second single takes or video editing and extension, 2.5 is the only Seedance tier that does it.
cURL
curl -X POST https://api.apimodels.app/v1/video/generations \
-H "Authorization: Bearer $APIMODELS_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "dreamina-seedance-2-0-mini",
"prompt": "A paper boat drifts down a rain-soaked street at dusk, slow dolly follow",
"resolution": "480p",
"duration": 8,
"aspect_ratio": "16:9"
}'
# then poll GET /v1/video/generations?task_id=...Python
import requests, time
H = {"Authorization": "Bearer YOUR_APIMODELS_KEY"}
r = requests.post("https://api.apimodels.app/v1/video/generations", headers=H, json={
"model": "seedance-2.0-mini", # accepts real-person reference photos as URLs
"prompt": "The woman in image 1 turns to camera and smiles",
"reference_image_urls": ["https://example.com/portrait.jpg"],
"resolution": "720p", "duration": 5,
}).json()
task = r["data"]["taskId"]
while True:
s = requests.get(f"https://api.apimodels.app/v1/video/generations?task_id={task}", headers=H).json()["data"]
if s["state"] in ("completed", "failed"):
print(s); break
time.sleep(5)At BytePlus list prices, Seedance 2.5 is $0.103 per second at 480p, $0.231 at 720p and $0.569 at 1080p; Seedance 2.0 is $0.07, $0.15 and $0.37; 2.0 Fast $0.06 and $0.12; 2.0 Mini $0.04 and $0.08. On APIMODELS the most affordable tier is Dreamina Seedance 2.0 Mini at $0.0225 per second at 480p.
Only on Dreamina Seedance 2.0 Mini, which is 44% below list. On 2.5, 2.0, 2.0 Fast and the seedance-2.0-mini line we are 10–33% above BytePlus list; on 2.5 and 2.0 we are still below fal, Kie.ai and EvoLink, while on Fast and Mini Kie.ai is more affordable than us. What you get for the difference is no enterprise sign-up, real-person references accepted on the seedance-2.0 line, and one balance for every other model.
Yes. With a reference video the per-token rate drops, but the input seconds are counted too, since tokens are (input + output seconds) × resolution. Editing or extending a clip of the same length as the output therefore usually costs more than generating it fresh.
The official Ark and BytePlus lines reject a reference showing an identifiable real person unless it is first registered in the asset library, and the Dreamina line on APIMODELS rejects it outright. The seedance-2.0, seedance-2.0-fast and seedance-2.0-mini models on APIMODELS accept a real-person photo as a plain URL and register it automatically when the upstream refuses.
