Best video generation model in the market.
30 seconds. One generation. Full creative control. Build cinematic AI video from text, images, video and audio references—with native sound and precise direction across every second.
Motion and audio are generated together across a full 30-second sequence — so pacing, atmosphere and on-screen events stay in sync.
One generation runs from 4 to 30 seconds, with continuity across connected shots.
Dialogue, ambience and sound sync with the visuals — generated in the same request.
Direct exactly what happens inside a specific time range, without breaking the scene.
Three capabilities move Seedance 2.5 beyond short single shots — longer connected scenes, deeper multimodal references and precise timestamp direction.
One generation carries connected shots from establishing beat to ending — up to 30 seconds with continuous motion and sound.
Feed up to 50 image, video and audio references and let the model fuse their subjects, movement, style and sound into one result.
Direct exactly what happens inside a specific time range — refine a character, action, camera move or story beat without breaking continuity.
Arrange connected shots, preserve subjects across changes, and generate sound with the image inside one coherent 30-second sequence.
Establish the setting, move through the action and land the ending inside one generation, with visual and audio continuity across connected shots.
Guide the generation with multimodal references and timestamp direction — while preserving continuity around what matters.
Seedance 2.5 can understand up to 50 multimodal references—images, video and audio—then combine their subjects, movement, style and sound into a unified result.
Describe what should happen inside a specific time range. Refine a character, action, camera move or story beat while preserving continuity around it.
Use the same asynchronous Seedance 2.5 endpoint for text-to-video, image-to-video and multimodal reference-to-video jobs.
Send prompt plus text, image, video and audio references in one JSON request.
Get a taskId back and track the job while rendering runs in the background.
Retrieve the finished MP4 through polling or a callback.
curl -X POST 'https://api.newportai.com/api/async/seedance_2.5' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer {{key}}' \
-d '{
"model": "seedance-2.5",
"prompt": "A girl dancing in a park on a sunny day with trees and benches in the background",
"resolution": "480p",
"ratio": "adaptive",
"duration": 8,
"seed": -1,
"generateAudio": false
}'const response = await fetch('https://api.newportai.com/api/async/seedance_2.5', {
method: 'POST',
headers: {
'Authorization': `Bearer ${process.env.DREAMAPI_KEY}`,
'Content-Type': 'application/json'
},
body: JSON.stringify({
model: 'seedance-2.5',
prompt: 'A girl dancing in a park on a sunny day with trees and benches in the background',
resolution: '480p',
ratio: 'adaptive',
duration: 8,
seed: -1,
generateAudio: false
})
});
const { data: { taskId } } = await response.json();HttpClient client = HttpClient.newHttpClient();
HttpRequest request = HttpRequest.newBuilder()
.uri(URI.create("https://api.newportai.com/api/async/seedance_2.5"))
.header("Content-Type", "application/json")
.header("Authorization", "Bearer {{key}}")
.POST(HttpRequest.BodyPublishers.ofString("{\"model\":\"seedance-2.5\",\"prompt\":\"A girl dancing in a park on a sunny day with trees and benches in the background\",\"resolution\":\"480p\",\"ratio\":\"adaptive\",\"duration\":8,\"seed\":-1,\"generateAudio\":false}"))
.build();
HttpResponse<String> res = client.send(request, HttpResponse.BodyHandlers.ofString());import requests
response = requests.post(
"https://api.newportai.com/api/async/seedance_2.5",
headers={
"Content-Type": "application/json",
"Authorization": "Bearer {{key}}"
},
json={
"model": "seedance-2.5",
"prompt": "A girl dancing in a park on a sunny day with trees and benches in the background",
"resolution": "480p",
"ratio": "adaptive",
"duration": 8,
"seed": -1,
"generateAudio": False
}
)
print(response.json())A concise view of the capabilities that change the production workflow.
One credit pool for every model on DreamAPI — video, image, avatar and editing. Buy a pack, then spend credits at the per-model rates below.
Credit packs are purchased on the DreamAPI pricing page.
Credits are calculated per generated second.
Billed per generated second — pay only for what you create.
Seedance 2.5 is ByteDance's multimodal AI video model for longer, controlled audio-video sequences.
A single generation can run from 4 to 30 seconds.
Yes. Native audio generation can be enabled in the same request.
DreamAPI provides 480p, 720p and 1080p output.
Submit an asynchronous request, receive a taskId, then retrieve the MP4 through polling or a callback.
Rates vary by resolution and active promotion. See the pricing table above for current output and reference-video credit rates.
From first reference to final frame—build longer, more controlled AI video through one production-ready API.