seedance-2-5 Edit or extend source video with Seedance 2.5, or create 4–30 second clips from text and multimodal references with native audio.
No GPU. No setup. Cancel anytime.
Seedance 2.5 is ByteDance's mainline video generation model, served through BytePlus ModelArk. It renders 4 to 30 second clips at a fixed 24 fps and generates native audio in the same pass.
Its Reference to Video workspace now exposes three explicit modes. Create makes a new clip from loose references. Edit transforms @Video1 while preserving everything you do not ask to change; the output follows the source video's duration and aspect. Extend continues @Video1 after its ending; the aspect follows the source while you choose a new 4 to 30 second continuation.
Seedance 2.5 accepts up to 30 reference images, 10 reference videos, and 10 reference audio clips — 50 media parts total. Edit and Extend require a source video as @Video1, and you can still use the wider reference stack to guide identity, environment, motion, or sound.
Seedance 2.5 is an addition to the Seedance line, not a replacement. Seedance 2.0 stays the default: 2.5 doubles the maximum clip length and widens reference input, but it tops out at 720p, while 2.0 also renders 1080p and 4K. Pick 2.5 for length and references, and 2.0 when resolution decides the shot.
Choose the reference mode first, then make @Video1's role explicit in the prompt.
In Edit and Extend, the first attached video is the source or continuation anchor. Edit inherits its duration and aspect; Extend inherits its aspect and uses the continuation duration you select.
The mode is fixed before generation so the same reference video cannot be ambiguously interpreted after billing.
| Mode | Required input | Output behavior |
|---|---|---|
| Create | At least one loose image, video, or audio reference | A new 4–30 second clip guided by the attached media |
| Edit | @Video1 as the source; optional supplemental references | Transforms the source; duration and aspect follow @Video1 |
| Extend | @Video1 as the continuation anchor | Continues after the ending; aspect follows @Video1, with a new 4–30 second duration |
Seedance 2.5 is a pay-as-you-go Premium Spark vendor model. Every clip uses purchased Spark, including for Sogni Unlimited members. Unlimited saves 5%; Unlimited Pro saves 10%.
| Configuration | Spark | USD |
|---|---|---|
| 5 s · 480p | 134 Spark | $0.67 |
| 30 s · 480p | 802 Spark | $4.01 |
| 5 s · 720p | 300 Spark | $1.50 |
| 10 s · 720p | 601 Spark | $3.00 |
| 30 s · 720p | 1803 Spark | $9.01 |
Per-second Premium Spark rates: $0.300456 per second at 720p and $0.133640 per second at 480p. Totals are billed from those exact rates.
Jobs that supply a reference video — video editing and video extension — use a lower vendor rate, so they cost less per second than text-to-video.
One Sogni API key reaches every model in the catalog — call Seedance 2.5 with the exact model id.
import { SogniClient } from '@sogni-ai/sogni-client';
const client = await SogniClient.createInstance({
appId: crypto.randomUUID(),
apiKey: process.env.SOGNI_API_KEY,
network: 'fast',
});
const project = await client.projects.create({
type: 'video',
modelId: 'seedance-2-5',
positivePrompt: 'a slothicorn surfing a wave of liquid paint, slow push-in, cinematic',
numberOfMedia: 1,
duration: 5,
});
const [url] = await project.waitForCompletion();
console.log(url); // result link — download within 24h curl https://api.sogni.ai/v1/creative-agent/workflows \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $SOGNI_API_KEY" \
-d '{
"input": {
"title": "First render",
"steps": [{
"id": "step1",
"toolName": "generate_video",
"arguments": { "prompt": "a slothicorn surfing a wave of liquid paint, slow push-in, cinematic", "videoModel": "seedance2-5", "duration": 5 }
}]
},
"confirm_cost": true
}' REST workflow alias: seedance2-5. Pin the exact id with the SDK. Full reference at docs.sogni.ai.
One 15-second motion-driven mascot edit became the control for a second 30-second edit: first the pink Sogni sloth replaces the source dancer, then the same mascot performs the choreography as a green-screen partner beside a live-action creator. Play both with sound and compare identity, plush body volume, motion fidelity, and beat synchronization.
Seedance 2.5 is a third-party frontier model. It uses purchased Spark for every clip; Unlimited plans discount those pay-as-you-go rates instead of covering them as fair-use renders.
Vendor-backed renders are billed with Spark even when your account has an active Unlimited subscription.
Unlimited saves 5%; Unlimited Pro saves 10%.
Launch Seedance 2.5 from the Sogni app or call the exact model id through the API.
Use premium vendor models alongside Sogni Supernet image, video, music, and language models.
Seedance 2.0 remains the default. Choose Seedance 2.5 when you need a clip longer than 15 seconds — it runs 4 to 30 seconds — or more reference material, up to 30 images, 10 videos, and 10 audio clips against 9, 3, and 3 on Seedance 2.0. Choose Seedance 2.0 when resolution decides it: Seedance 2.5 renders 480p and 720p only, while 2.0 also offers 1080p and 4K.
Text on its own; a first frame, or a first-and-last-frame pair, for image-to-video; and up to 30 loose images, 10 videos, and 10 audio clips for Create, Edit, and Extend reference modes, with 50 media parts total. Audio-only reference creation is supported.
Both require a source video as @Video1 and inherit its aspect ratio. Edit transforms the existing timeline and automatically follows @Video1's duration, up to 30 seconds. Extend starts after @Video1 ends and uses a new continuation duration you choose from 4 to 30 seconds.
Set seedanceTaskType to reference, edit, or extend on a seedance-2-5 video project. Edit and Extend require at least one reference video; the first attached video is @Video1. Omit seedanceTaskType for first-frame and first-plus-last-frame generation.
Yes. Seedance 2.5 generates native audio as part of the same pass, at a fixed 24 fps.
No. Seedance 2.5 renders 480p and 720p only. For 1080p or 4K, use Seedance 2.0.
Not currently. Seedance 2.5 is not an open-weight model, so Sogni cannot run it directly on Supernet worker GPUs. We provide it through the third-party BytePlus ModelArk API, which charges Sogni per render, so Seedance 2.5 must remain pay-as-you-go. Unlimited members save 5% and Unlimited Pro members save 10% on those rates. ByteDance has not provided a forecast for whether or when Seedance 2.5 will become open weight; if it becomes available to run on Sogni worker nodes, we will add it to Unlimited. Until then, MiniMax H3 and Wan 2.2 are excellent open-weight video alternatives included with Unlimited.
Seedance 2.5 is pay-as-you-go Premium Spark. It is not covered as an Unlimited fair-use render; every clip uses purchased Spark. Unlimited saves 5%; Unlimited Pro saves 10%. Prices start at 134 Spark ($0.67) per clip.
Create in the app, or build with the API. Your call.