Seedance 2.5 API: Price per Second, Model IDs and Python Example

Short answer

ByteDance Seed released Seedance 2.5 on July 31, 2026 (launch post), and TeamoRouter supports Seedance 2.5. It's ByteDance's video model for long single takes, with sound generated together with the picture.

• What it does: up to 30 seconds per generation (Seedance 2.0: 15), extension over several rounds, up to 30 images, 10 videos and 10 audio clips as references, and edits pinned to timestamps.
• Official API price: BytePlus ModelArk charges $10.70 per 1M tokens at 480p and 720p, and $11.70 at 1080p. At 720p that's about $0.23 per second, or $1.16 for a 5-second clip (BytePlus pricing).
• On TeamoRouter: three model IDs (text, image and reference to video), 4–30 seconds at 480p, 720p or 1080p, through the OpenAI-compatible /v1/videos API. One key and a prepaid balance from $5. No BytePlus account needed.

New accounts start with a $0 balance — top up before your first request. Top up from $5. If a request returns 402 (400 on the OpenAI format) with "wallet balance is insufficient", your balance is empty — your base URL and key are fine. Top up and retry.

Get a key and generate video with Seedance 2.5

What Seedance 2.5 can do

From ByteDance Seed's launch post and BytePlus's Seedance 2.5 tutorial:

• Stories in one clip. Within 30 seconds the model can plan several connected shots, so a scene can go from setup to ending without stitching clips together.
• Better picture and sound. ByteDance says image, audio and motion quality improved, with a more natural look than typical AI video.
• 11 languages. BytePlus lists prompts and audio in Chinese, English, Spanish, Indonesian, Malay, Thai, Arabic, Portuguese, Vietnamese, Japanese and Korean.
• Where ByteDance offers it: the Jimeng AI and Doubao Pro apps, and its own API on BytePlus ModelArk (model dreamina-seedance-2-5-260628) and on Volcengine Ark in China.

Call Seedance 2.5 through TeamoRouter

Python with the OpenAI SDK — create a job, poll it, download the MP4:

import os, time
from openai import OpenAI

client = OpenAI(api_key=os.environ["TEAMO_API_KEY"], base_url="https://api.teamorouter.cn/v1")
video = client.videos.create(
    model="seedance-2-5-text-to-video-basic",
    prompt="One continuous shot: a street violinist starts playing in the rain, the camera circles him as a crowd gathers, sound of rain and violin",
    seconds="15",
    size="1280x720",
)
while video.status in ("queued", "in_progress"):
    time.sleep(10)
    video = client.videos.retrieve(video.id)
if video.status == "failed":
    raise RuntimeError(video.error)
client.videos.download_content(video.id).write_to_file("seedance.mp4")

• seconds is a whole number from 4 to 30, sent as a string.
• size picks resolution and aspect ratio: 854x480 is 480p 16:9, 1280x720 is 720p 16:9, 1080x1920 is 1080p 9:16. Without size, the model's first listed resolution and aspect ratio are used.

The same job with curl:

curl https://api.teamorouter.cn/v1/videos \
  -H "Authorization: Bearer $TEAMO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"seedance-2-5-text-to-video-basic","prompt":"A paper lantern rises over a busy night market, slow tilt up, crowd noise","seconds":"10","size":"1920x1080"}'

The response has an id. Poll GET /v1/videos/{id} until status is completed, then download GET /v1/videos/{id}/content. Finished videos are kept for seven days.

Image to video — pass your picture in input_reference:

video = client.videos.create(
    model="seedance-2-5-image-to-video-basic",
    prompt="The woman in the photo turns to the camera and smiles, wind in her hair",
    seconds="8",
    input_reference={"image_url": "https://example.com/portrait.jpg"},
)

image_url can be a public URL or a base64 data URL. Poll and download the same way.

• GET /v1/models?type=video lists each Seedance 2.5 model ID with the durations, resolutions and aspect ratios it accepts. Full reference: video generation API.
• ByteDance's own API also runs video editing and extension tasks. On TeamoRouter, /v1/videos runs text-to-video, image-to-video and reference-to-video jobs.
• Prefer a UI? The video workbench uses the same balance.

Which model ID for which job

All three take 4–30 seconds at 480p, 720p or 1080p. Text to video and reference to video take 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16.

• Just an idea, no assets: text to video. Describe the shots in order. ByteDance's own example prompts pin events to time ranges, such as "0–5 seconds: …".
• You already have the opening frame, such as a product shot or a storyboard frame: image to video.
• A character or product must look the same in every shot, but it doesn't have to be the first frame: reference to video. For several reference images, use the extension field reference_image_urls (in the OpenAI SDK, pass it in extra_body).

Seedance 2.5 API price

ByteDance publishes the API price on BytePlus ModelArk in US dollars per 1M tokens:

BytePlus estimates tokens as (input video seconds + output seconds) × width × height × frame rate ÷ 1024. Its own examples for a 16:9 clip without input video:

The 30-second column is the per-second price × 30.

• BytePlus charges only for videos that finish. A job that fails, for example in content moderation, is free.
• When the input includes a video, its length counts toward tokens and a minimum token charge applies.
• Before you can turn the model on, BytePlus asks for one of these: a balance over $30, an AI Savings Plan of $30 or more, or a Seedance 2.5 resource pack (tutorial).

TeamoRouter's current Seedance 2.5 price is on the Seedance 2.5 model page and the pricing page.

Paying for it

There's no subscription. Top up from $5 by card, and each video job is charged to your balance. The same key and balance also call DeepSeek, Kimi and other text models.

FAQ

Is there an official Seedance 2.5 API?

Yes. ByteDance sells it on BytePlus ModelArk as dreamina-seedance-2-5-260628, and on Volcengine Ark in China. BytePlus asks for a balance over $30, a $30+ savings plan or a resource pack before you can turn the model on. Through TeamoRouter you start from $5 with one key.

How long can a Seedance 2.5 video be?

4 to 30 seconds per job. ByteDance's apps and API can also extend a clip over several rounds.

Is this the same Seedance 2.5 as in Jimeng and Doubao?

Yes. ByteDance Seed's launch post says Seedance 2.5 is rolling out in Jimeng AI and Doubao Pro. Here you call the model through an API by model ID and pay per job, with no app membership.

Does Seedance 2.5 make sound?

Yes. ByteDance generates audio and video together. BytePlus lists 11 languages for prompts and audio, including English, Chinese, Spanish and Japanese.

Related

• Seedance 2.5 model page: IDs and current price
• Video generation API
• All video models