# Seedance 2.5

> Seedance 2.5 is ByteDance’s flagship multimodal video generation model, producing continuous clips from 4 to 30 whole seconds across 480P, 720P, 1080P, and native 4K UHD, with controllable native audio generation and reproducible random seeds.

Canonical: https://routemux.com/bytedance/seedance-2.5
Provider: ByteDance

## Key facts

| Field | Value |
| --- | --- |
| Model ID | bytedance/seedance-2.5 |
| Official model ID | seedance-2.5 |
| Input modalities | text, image |
| Output modalities | video |
| Capabilities | Video, Text To Video, Image To Video, Reference Audio, Reference Image, Reference Video, First Last Frame |
| Upstream release | 2026-07-31 |
| Availability | available |

## Pricing

- RouteMux: $0.1201 per second (from — tiered by duration/resolution)
- Official: $0.1201 per second
  - 480P: $0.1201 per second (official $0.1201)
  - 720P: $0.27 per second (official $0.27)
  - 1080P: $0.6683 per second (official $0.6683)
  - 4K: $1.0761 per second (official $1.0761)

Billing is prepaid wallet, charged per successful request only.

## API access

Base URL: https://api.routemux.com

Supported protocols:

- openai_sora

## What is Seedance 2.5

Seedance 2.5 is ByteDance’s flagship multimodal video generation model, producing continuous clips from 4 to 30 whole seconds across 480P, 720P, 1080P, and native 4K UHD, with controllable native audio generation and reproducible random seeds.

Three workflows are supported on RouteMux: text-to-video from a prompt alone; image-to-video from a first frame with an optional last frame; or large-scale multimodal references — up to 30 reference images, 10 reference videos (2–15 seconds each, up to 30 seconds total), and 10 reference audio clips (2–15 seconds each, up to 30 seconds total; audio references require at least one reference image or video in the same request), up to 50 input files total. First/last frames and reference media cannot be mixed.

Aspect ratio accepts 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, or adaptive (defaulting to 16:9 for text-to-video and adaptive when visual media is supplied; at 4K with first/last frames, adaptive is enforced automatically). Pass audio (or generate_audio) as true or false to toggle sound at the same price, and pass seed (-1 or 0..2147483647) for repeatable framing.

Without reference video, jobs bill per output second at the selected resolution’s rate. With reference_videos, the entire job switches to the lower reference-video per-second rate for (output_seconds + reference_video_seconds), while reference images and audio add no surcharge. Runs asynchronously via POST /v1/videos; failed jobs are not billed.

## Key capabilities

- **480P up to native 4K** — Four resolution tiers from fast 480P previews to 4K UHD deliverables.
- **4–30 s + audio & seed** — Whole-second lengths with optional native audio toggle and seed control.
- **Up to 50 reference files** — Combine up to 30 images, 10 videos, and 10 audio clips in one request.
- **Async Sora-compatible API** — Unified POST /v1/videos submission and GET /v1/videos/{id} polling.

## Sources

- [ByteDance Seed — Official Research Page](https://seed.bytedance.com)
- [Volcengine Ark Video Generation Docs](https://www.volcengine.com/docs/82379)

_Content reviewed by the RouteMux team on 2026-11-19._
