Seedance 2.0 API Pricing: 2.0 vs Fast vs Mini vs 2.5

On Orux AI, the Seedance 2.0 API costs 61.5 Credits ($0.3075) per second at 720p, so a 5-second 720p clip is 307.5 Credits ($1.54). The cheapest tier, Seedance 2.0 Mini, starts at 12.59 Credits ($0.063) per second at 480p, and Seedance 2.5 tops out at 237 Credits ($1.185) per second at 1080p. Price depends only on the model, resolution and duration; turning audio on does not change it.

This guide lays out all four tiers per second and per clip, puts them next to the official and third-party prices we could find (including where Orux AI is more expensive), and gives a working Python script for text-to-video and first-frame image-to-video. Prices as of Oct 29, 2026.

Seedance price per second: all tiers

Orux AI prices in Credits per second of output video (1 Credit = $0.005):

Swipe horizontally to view the full table.

The jump from 720p to 1080p is large: on Seedance 2.0 it is 2.5 times the per-second price, and on 2.5 it is 2.5 times as well. 4K is only offered on Seedance 2.0 and costs about five times its 720p rate.

What a 5-second and 10-second clip costs

Most people budget per clip, not per second. Same prices, multiplied out:

Swipe horizontally to view the full table.

A 30-second single clip is only possible on Seedance 2.5; at 720p it costs 2,835 Credits ($14.18). For reference, Kling 4.0 at 720p costs 42.75 Credits per second on Orux AI, which sits between Seedance 2.0 Fast and Seedance 2.0.

Official and third-party prices

ByteDance sells Seedance in China through Volcengine Ark, priced in CNY. We did not convert these to USD, because exchange rates and account terms differ. At the time of writing, Volcengine's pricing page lists approximately these prices for a 5-second 16:9 clip without video input:

Swipe horizontally to view the full table.

Volcengine prices video generation by tokens, so the per-clip figure shifts with resolution, ratio and duration. Check the page for the exact calculation before relying on these numbers.

fal, a third-party inference platform, published per-second prices in a Seedance comparison article in August 2026:

Swipe horizontally to view the full table.

So Orux AI is not cheaper than fal across the board. At 720p the two are effectively level, Fast is slightly cheaper on Orux AI, and Seedance 2.0 at 1080p costs more on Orux AI. fal's article dates its rates to August 10, 2026, and its own per-second figures do not exactly match its stated token formula, so confirm current prices on fal before you compare.

What Orux AI adds is operational: the same key and balance for Seedance, Kling, Wan, image and chat models; one submit-and-poll interface across all of them; and no charge for tasks that fail or expire.

Which tier for which job

Swipe horizontally to view the full table.

The table is about price and limits. We have not run a controlled quality comparison between tiers, so the only safe rule is to test the cheap tier first and move up only when the result is not good enough for the job. A single 5-second test at Mini 480p costs less than a quarter of the same test on 2.0 at 720p.

Python: text-to-video and first-frame image-to-video

Video generation on Orux AI is asynchronous: submit to POST /api/v1/tasks, get a task_id back with HTTP 202, then poll GET /api/v1/tasks/{task_id} until the status is SUCCESS, FAILED or EXPIRED. Only SUCCESS is billed, and there is no callback option.

import os, time, json, requests

BASE = "https://orux.top/api/v1"
HEADERS = {"Authorization": f"Bearer {os.environ['ORUX_API_KEY']}"}

# Credits per second on Orux AI, checked 2026-10-29
PER_SECOND = {
    "seedance-2.0-mini": {"480p": 12.586, "720p": 28.318},
    "seedance-2.0-fast": {"480p": 20.2468, "720p": 45.5548},
    "seedance-2.0":      {"480p": 28.5, "720p": 61.5, "1080p": 153, "4k": 312},
    "seedance-2.5":      {"480p": 42, "720p": 94.5, "1080p": 237},
}

def estimate(model: str, resolution: str, duration: int) -> float:
    return round(PER_SECOND[model][resolution] * duration, 6)

def submit(model: str, payload: dict) -> str:
    r = requests.post(f"{BASE}/tasks", headers=HEADERS,
                      json={"model": model, "input": payload}, timeout=60)
    if r.status_code == 402:
        raise RuntimeError("Insufficient balance: top up (not an auth error)")
    r.raise_for_status()
    task_id = r.json()["task_id"]
    with open("video_tasks.jsonl", "a") as f:   # persist before polling
        f.write(json.dumps({"task_id": task_id, "model": model}) + "\n")
    return task_id

def wait_and_save(task_id: str, path: str, max_wait: int = 1800) -> None:
    deadline = time.time() + max_wait
    while time.time() < deadline:
        time.sleep(10)                           # videos take minutes; poll gently
        t = requests.get(f"{BASE}/tasks/{task_id}", headers=HEADERS, timeout=30).json()
        if t["status"] == "SUCCESS":
            with open(path, "wb") as f:
                f.write(requests.get(t["result"][0], timeout=300).content)
            print("saved", path, "| link valid for", t.get("result_expires_in"))
            return
        if t["status"] in ("FAILED", "EXPIRED"):
            raise RuntimeError(t.get("error") or t["status"])   # not charged
    raise TimeoutError(f"{task_id} still running; resume polling, do not resubmit")

# 1) Text-to-video: cheap draft on Mini
draft = {
    "prompt": "A slow cinematic camera push towards a ceramic mug on a wooden desk "
              "in soft morning light. No people, no text.",
    "resolution": "480p",
    "duration": 5,
    "aspect_ratio": "16:9",
    "generate_audio": False,
}
print("estimate:", estimate("seedance-2.0-mini", "480p", 5), "Credits")
wait_and_save(submit("seedance-2.0-mini", draft), "draft.mp4")

# 2) Image-to-video: animate from a first frame on Seedance 2.0 at 720p
i2v = {
    "prompt": "The steam rises and drifts left; the camera slowly orbits the mug.",
    "first_frame_url": "https://example.com/mug-first-frame.jpg",
    "resolution": "720p",
    "duration": 5,
    "aspect_ratio": "adaptive",
}
print("estimate:", estimate("seedance-2.0", "720p", 5), "Credits")
wait_and_save(submit("seedance-2.0", i2v), "final.mp4")

Notes on running this in production:

• Persist the task_id before polling, as the script does. If the process dies, resume polling the same id; resubmitting would generate and bill a second video.
• Result URLs are temporary. The task response includes result_expires_in; download and store the file yourself as soon as the task succeeds.
• A 402 response means the balance is too low, not that the key is wrong. Route it to your top-up alert rather than your auth alert.
• On 429, back off exponentially with jitter.
• 4K on Seedance 2.0 uses the lowercase value 4k in resolution.

Audio, aspect ratios and duration limits

What each tier accepts on Orux AI:

Swipe horizontally to view the full table.

Audio generation is on by default on every tier. Orux AI's published Seedance prices vary only by model, resolution and duration, so a clip with audio costs the same as one without. Turn it off with "generate_audio": false when you plan to add your own soundtrack, simply to avoid unwanted sound.

return_last_frame on Seedance 2.5 returns the final frame of the clip, which you can pass as first_frame_url to the next shot to chain clips together. That is one way to build sequences longer than 15 seconds on the cheaper tiers.

One prompt on four tiers

The Orux model pages for all four tiers share an existing example, "Morning light", generated from the same prompt with identical settings: 4 seconds, 480p, 16:9, audio off. The saved files run 4.04 seconds; the 2.0, Fast and Mini files are 864×496 and the 2.5 file is 854×480. Two of them:

At these settings the four clips cost 50.34 Credits on Mini, 80.99 on Fast, 114 on 2.0 and 168 on 2.5. Watch them side by side on the model pages and judge whether the difference matters for your use case; that comparison is worth more than any general claim about tiers.

For per-second prices across Kling, Wan and Vidu as well, see our cheapest AI video generation API comparison, and for Kling's parameters and pricing, the Kling 4.0 API guide.

FAQ

Q.How much does Seedance 2.0 cost per second?

On Orux AI, Seedance 2.0 costs 28.5 Credits ($0.1425) per second at 480p, 61.5 ($0.3075) at 720p, 153 ($0.765) at 1080p and 312 ($1.56) at 4K, as of Oct 29, 2026. A 5-second 720p clip is 307.5 Credits ($1.54).

Q.What is the difference between Seedance 2.0 Mini and Fast?

On Orux AI both offer 480p and 720p, 4–15 second clips and the same inputs. Mini costs 12.59 Credits per second at 480p versus 20.25 for Fast, about 62% of the price. We have not measured a quality or speed difference, so test both on your prompt before choosing.

Q.Is Seedance 2.5 available via API?

Yes. On Orux AI it is the seedance-2.5 model on the same tasks endpoint, at 42, 94.5 and 237 Credits per second for 480p, 720p and 1080p. It is the only tier with clips up to 30 seconds and up to 30 reference images.

Q.Does audio cost extra?

No. Orux AI's Seedance prices depend only on model, resolution and duration, and generate_audio is on by default. A clip with generated audio costs the same as a silent one. fal's August 2026 article also describes native audio as included at no extra charge on its platform.

Q.What is the maximum video length?

Seedance 2.0, 2.0 Fast and 2.0 Mini accept 4 to 15 seconds per clip. Seedance 2.5 accepts 4 to 30 seconds. For longer sequences, generate several clips and use the last frame of one as the first frame of the next.

Q.Am I charged if a video fails?

No. Video tasks are billed only when they reach SUCCESS. Tasks that end as FAILED or EXPIRED cost nothing. Billing is from a prepaid balance with no monthly fee, so the estimate you calculate before submitting is the most you pay for that clip.

Sources and price check

Prices checked on 2026-10-29. Orux AI rates and parameters: the Seedance 2.0, 2.0 Fast, 2.0 Mini and 2.5 model pages. Volcengine Ark prices (CNY): Volcengine model pricing. fal prices: Seedance 2.5 vs Seedance 2.0 on fal (rates dated Aug 10, 2026). Re-check all three before budgeting.