Short answer, checked on 17 July 2026: the Seedance 2.5 API is NOT open yet. Many sites report it went live on BytePlus on 16 July — we tested that directly. Querying the BytePlus ModelArk model list with our production key returns 52 models, and the Seedance family stops at the Seedance 2.0 tiers; every Seedance 2.5 model id we probed returns "InvalidEndpointOrModel.NotFound". So as of today nobody can call it through us, and we would rather say so than sell you a page that 404s on generate. We re-check regularly and will list it the day it opens.
What Seedance 2.5 is: ByteDance unveiled it on 23 June 2026 at its Volcano Engine FORCE conference, skipping several intermediate versions. Reported specs are a native 30-second single-shot clip in one pass (roughly double Seedance 2.0), native 4K at 10-bit colour rather than an upscale, up to 50 multimodal reference assets at once (images, video clips, audio tracks and 3D white-box models), localised scene editing so you can swap a product or background without re-rolling the whole clip, and audio co-processed in the same latent space as the visuals so dialogue and effects stay locked to the action. Treat these as ByteDance's announced figures — we will publish measured numbers and a real per-second price once we can actually run it.
What you can use right now: Seedance 2.0 on the direct official Volcengine Ark channel, from $0.092/s — the same family, already generating native synced audio, up to 9 reference images plus a reference video and audio, 480p/720p/1080p, 4-15 seconds. No Volcengine account and no enterprise verification: sign up, create an API key, call model "seedance-2.0". The sample below is a real, working call. When 2.5 opens, the same key and the same endpoint will reach it — you will only change the model string.
cURL
# Seedance 2.5 is not callable yet (checked 2026-07-17). This is Seedance 2.0,
# which works today — when 2.5 opens, only the "model" value changes.
curl https://apimodels.app/api/v1/video/generations \
-H "Authorization: Bearer $APIMODELS_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "seedance-2.0",
"prompt": "a hummingbird hovering near a bright flower, slow motion",
"resolution": "720p",
"duration": 5
}'
# -> { "data": { "taskId": "..." } }, then poll:
# curl "https://apimodels.app/api/v1/video/generations?task_id=..." -H "Authorization: Bearer $APIMODELS_API_KEY"Python
import requests, time
H = {"Authorization": "Bearer YOUR_KEY", "Content-Type": "application/json"}
B = "https://apimodels.app/api/v1/video/generations"
# Seedance 2.5 is not open yet — this calls Seedance 2.0 (available today).
task = requests.post(B, headers=H, json={
"model": "seedance-2.0", # swap to the 2.5 id once it launches
"prompt": "a hummingbird hovering near a bright flower, slow motion",
"resolution": "720p", "duration": 5,
}).json()["data"]["taskId"]
while True:
r = requests.get(B, headers=H, params={"task_id": task}).json()["data"]
if r["state"] in ("completed", "failed"):
print(r); break
time.sleep(5)Not as of 17 July 2026. We queried the BytePlus ModelArk model list with our production key: 52 models are exposed and the Seedance family stops at the Seedance 2.0 tiers (dreamina-seedance-2-0, -fast, -mini). Every Seedance 2.5 id we probed returns "InvalidEndpointOrModel.NotFound". Reports of a 16 July API launch do not match what the API actually returns. We re-check regularly and will list it the day it opens.
It was unveiled on 23 June 2026 at ByteDance's Volcano Engine FORCE conference, with an early-July launch window widely reported and an API date of 16 July cited by BytePlus coverage. That API date has not materialised for third-party accounts like ours. ByteDance has not published a firm public API date beyond that, so we are not going to guess one — follow this page and we will update it the moment the model appears in the live model list.
By ByteDance's announced figures: 30-second single-shot clips (Seedance 2.0 tops out at 15s), native 4K at 10-bit instead of Seedance 2.0's 1080p ceiling on the official channel, up to 50 multimodal references (Seedance 2.0 takes up to 9 images plus a reference video and audio), plus localised scene editing and audio co-processed in the same latent space. Seedance 2.0 already generates native synced audio, so that part is an improvement in tightness rather than a new capability.
Nobody can call it via a third-party API yet. When it opens, using it through apimodels.app needs no Volcengine account and no enterprise verification — you sign up, create one API key, and pass the model name; the same key already calls Seedance 2.0, VEO, Kling, Sora and the rest. If you have a workload waiting, build it against seedance-2.0 today: the endpoint, auth and polling are identical, so migrating is a one-string change.
Not announced, and we will not invent a number. For reference, our official-channel Seedance 2.0 pricing is per second: 480p $0.092/s, 720p $0.197/s, 1080p $0.492/s, with cheaper Fast ($0.071/s) and Mini ($0.044/s) tiers — charged only on success. Expect 4K and 30-second output to cost materially more per clip simply because there are far more pixels and seconds. We will publish the real rate when we can run it.