Blog

Guide

Seedance 2.0 API Guide: Mini, Fast, Standard and the ext Route

Which Seedance 2.0 version to call, a same-prompt test across three tiers, image-to-video and references, and working curl and Python.

Jun 23, 2026 7
Seedance 2.0 API Guide: Mini, Fast, Standard and the ext Route
Mini 720p / sec
$0.0586
Standard 720p / sec
$0.293
Clip length
4–15 s
On this page

TL;DR

  • Seedance 2.0 is ByteDance's video model. On HiAPI you can call it four ways: Mini, Fast, standard, and standard on the ext route. All four use the same request shape; only the model id or one route field changes.
  • Per second at 720p: Mini $0.0586, ext $0.134, Fast $0.1772, standard $0.293. Sound is on by default and costs nothing extra. Clips run 4–15 seconds.
  • Draft on Mini, ship on ext for 1080p or 4K, and use standard when you need a reference video. Open the Seedance 2.0 model page to try it; new accounts get 200 free credits.

Which version to call

Call How to select it Resolutions Reference video 720p per second
Mini "model": "seedance-2.0-mini" 480p, 720p Yes $0.0586
Fast "model": "seedance-2.0-fast" 480p, 720p Yes $0.1772
Standard "model": "seedance-2.0" 480p, 720p, 1080p, 4K Yes $0.293
ext route "model": "seedance-2.0" plus "route": "ext" 720p, 1080p, 4K No $0.134

Live HiAPI prices, checked October 19, 2026. Failed tasks aren't charged.

  • Prompt tuning and bulk drafts: Mini at 480p, $0.0272 per second.
  • 1080p or 4K finals: the ext route, $0.342 per second at 1080p and $0.694 at 4K, about half the standard rate.
  • Copying motion or camera from a reference video: standard, Fast or Mini. The ext route rejects reference_video_urls.

Every rate, including the cheaper per-second prices when you pass a reference video, is in Seedance 2.0 API pricing.

Same prompt on Mini, Fast and standard

720p, 5 seconds, 16:9, sound on, one run each:

Close-up in a sunlit corner cafe: a barista tilts a small steel pitcher and pours steamed milk into a white ceramic cup, a rosetta latte-art leaf blooming on the coffee surface. The camera starts tight on the cup and slowly pulls back to reveal the wooden counter and a window with morning street light. Soft hiss of the steam wand and gentle cafe chatter in the background. Warm natural color, shallow depth of field, smooth handheld feel.

Generate this prompt on Mini →

Mini, $0.293. The pour finishes in the first half, then the camera pulls back to the window-side table.

Fast, $0.886. The pour runs into the middle of the clip and the pull-back shows the hand on the pitcher.

Standard, $1.465. The pull-back goes furthest and plays out the "reveal the counter" part of the prompt most fully.

All three followed the prompt. Standard did the most with the camera move; Mini got the shot for a fifth of the price.

Two more standard clips

A product shot from one prompt, sound off:

A slow cinematic dolly-in on a sleek matte-black smartwatch resting on a brushed-concrete pedestal, soft diffused studio light, the screen wakes to show a glowing minimalist watch face, faint reflections move across the metal bezel, premium tech commercial look, calm and clean.

seedance-2.0, 5 s, 720p, 16:9, generate_audio: false, $1.465.

A mood shot for a page header:

seedance-2.0, 5 s, 720p, 16:9, $1.465.

curl: your first clip

Start on Mini at 480p for 5 seconds, $0.136:

curl -s -X POST https://api.hiapi.ai/v1/tasks \
  -H "Authorization: Bearer $HIAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2.0-mini",
    "input": {
      "prompt": "A slow dolly-in on a matte-black smartwatch on a concrete pedestal, soft studio light",
      "resolution": "480p",
      "duration": 5,
      "aspect_ratio": "16:9"
    }
  }'

You get a taskId back. Poll GET https://api.hiapi.ai/v1/tasks/{taskId} until data.status is success, then download data.output[0].url. The link expires, so save the file right away.

For the 1080p final on the ext route, change two things:

{
  "model": "seedance-2.0",
  "route": "ext",
  "input": {
    "prompt": "A slow dolly-in on a matte-black smartwatch on a concrete pedestal, soft studio light",
    "resolution": "1080p",
    "duration": 5,
    "aspect_ratio": "16:9"
  }
}

route sits at the top level, next to model, not inside input.

Python: draft on Mini, finish on ext

# pip install requests; export HIAPI_KEY=your-key first
import os, time, uuid, requests

API = "https://api.hiapi.ai/v1/tasks"
H = {"Authorization": f"Bearer {os.environ['HIAPI_KEY']}"}

def generate(model, inp, route=None, timeout_s=1200):
    body = {"model": model, "input": inp}
    if route:
        body["route"] = route
    # Reuse the same Idempotency-Key if you retry a submit, so a retry never bills twice
    r = requests.post(API, json=body, headers={**H, "Idempotency-Key": str(uuid.uuid4())}, timeout=30)
    res = r.json()
    if res.get("code") != 200:
        raise RuntimeError(res)  # 400s list every invalid field
    task_id = res["data"]["taskId"]
    deadline = time.time() + timeout_s
    while time.time() < deadline:
        d = requests.get(f"{API}/{task_id}", headers=H, timeout=30).json()["data"]
        if d["status"] == "success":
            return d["output"][0]["url"]
        if d["status"] == "fail":
            raise RuntimeError(d)
        time.sleep(10)
    raise TimeoutError(task_id)

shot = {"prompt": "A barista pours latte art, camera slowly pulls back to reveal the cafe",
        "duration": 5, "aspect_ratio": "9:16"}

draft = generate("seedance-2.0-mini", {**shot, "resolution": "480p"})            # $0.136
final = generate("seedance-2.0", {**shot, "resolution": "1080p"}, route="ext")   # $1.71

with open("final.mp4", "wb") as f:
    f.write(requests.get(final).content)

Image-to-video and references

Add inputs to the same request:

Field What it does Limits
first_frame_url Animates from this image No SVG
last_frame_url Sets the ending frame Needs first_frame_url
reference_image_urls Locks a character, product or style Up to 9 images, counting the frames
reference_video_urls Copies motion or camera work Up to 3; not on the ext route
reference_audio_urls Guides rhythm or voice Up to 3
{
  "model": "seedance-2.0",
  "input": {
    "prompt": "The product rotates slowly as a light sweep crosses the surface",
    "first_frame_url": "https://your-cdn.example.com/product.jpg",
    "aspect_ratio": "adaptive",
    "resolution": "720p",
    "duration": 5
  }
}

adaptive makes the clip follow the first frame's shape. With a reference video, billing counts only the seconds you generate, at the lower reference-video rate.

Parameters

Field Required Values Notes
prompt Yes 3–20,000 characters English and Chinese both work
aspect_ratio Yes on standard and ext 16:9 (default), 9:16, 1:1, 4:3, 3:4, 21:9, adaptive Set it on every call to be safe
duration Yes on Mini 4–15, default 5 Billed per second
resolution Yes on Mini Mini/Fast: 480p, 720p; standard: up to 4k; ext: 720p, 1080p, 4k Default 720p elsewhere
generate_audio No true (default) / false Same price either way
return_last_frame No true / false Standard only; returns the final frame as an image, handy for chaining clips

Errors you'll hit

All of these return HTTP 400 when you submit, so no task is created and nothing is billed:

  • aspect_ratio: missing required field: standard and ext need it.
  • resolution: missing required field or duration: missing required field: Mini needs both.
  • duration: minimum: got 3, want 4: clips start at 4 seconds.
  • resolution: value must be one of '480p', '720p': Mini and Fast stop at 720p. Use standard or ext for 1080p.
  • additional properties 'reference_video_urls' not allowed: you sent a reference video on the ext route. Drop route or the video.

FAQ

Does turning sound off save money?

No. Sound on or off costs the same on every version.

Can I write prompts in Chinese?

Yes. English and Chinese both work. Writing the camera move, the action and the sound as separate sentences gives the most consistent results.

How do I make vertical video?

Set aspect_ratio to 9:16. For image-to-video, adaptive follows the image.

Mini, Fast or standard for social clips?

Start on Mini 720p. If the motion or hands aren't right, run the same prompt on Fast and keep the better take. Use standard or ext only for the final cut.

More prompts: Seedance 2.0 ext prompt recipes and Seedance 2.0 Mini for short-form video. For the newer generation, see Seedance 2.5 vs 2.0.

Related articles

View all articles →

Start building with persistent AI artifacts

Get your API key and call image, video and audio models through one task-based API.

Read the API docs