TL;DR
Prices below are list prices as of November 14, 2026, for one 16:9 clip with no input video.
- Cheapest Seedance 2.5 API: BytePlus ModelArk, ByteDance's international cloud, at $10.70 per 1M video tokens for 480p and 720p and $11.7 for 1080p (BytePlus pricing). Replicate charges the same per second at 480p and 720p but has no 1080p option.
- One 10-second Seedance 2.5 clip on BytePlus: $1.03 at 480p, $2.31 at 720p and $5.69 at 1080p. At 720p the same clip costs $2.43 on Segmind, $2.59 on laozhang.ai, $3.15 on kie.ai, $3.50 on PiAPI, $3.60 on WaveSpeedAI and $4.62 on fal.ai.
- Cheapest Seedance of any version: Seedance 2.0 Mini, at $0.35 (480p) or $0.76 (720p) per 10 seconds on BytePlus, and $0.19 or $0.41 on kie.ai, the only reseller below BytePlus on Mini. Seedance 2.5 Flash becomes callable on BytePlus at 14:00 UTC+8 on November 15, 2026, at $0.67 (480p) or $1.51 (720p) per 10 seconds.
- Getting official access: BytePlus activates Seedance once you hold USD 30 in credits, a Savings Plan or a resource pack. No BytePlus page asks for a $1 million deposit.
- Stability: neither BytePlus nor the resellers below publish a Seedance success rate. What you can compare is failure billing (BytePlus charges only for successfully generated videos) and concurrency (3 tasks at a time on an individual BytePlus account, 10 on an enterprise account).
- What breaks the estimate: Seedance 2.5 defaults to
duration: -1and may return up to 30 seconds, input video seconds are billed too, and every take you throw away is a full charge.
Seedance 2.5 Cost per 10-Second Video by Provider and Resolution
As of November 14, 2026, one 10-second, 16:9 Seedance 2.5 video with no input video costs $1.03 at 480p, $2.31 at 720p and $5.69 at 1080p at BytePlus list price. Resellers charge between the same amount and twice that.
| Provider | 480p | 720p | 1080p | 720p vs BytePlus |
|---|---|---|---|---|
| BytePlus ModelArk (official) | $1.03 | $2.31 | $5.69 | 1.00× |
| Replicate | $1.03 | $2.31 | not offered | 1.00× |
| Segmind | $1.08 | $2.43 | $5.99 | 1.05× |
| laozhang.ai | $1.15 | $2.59 | $6.42 | 1.12× |
| kie.ai | $1.40 | $3.15 | $7.90 | 1.36× |
| PiAPI | $1.50 | $3.50 | $8.00 | 1.51× |
| WaveSpeedAI | $1.80 | $3.60 | $9.00 | 1.56× |
| fal.ai | $2.06 | $4.62 | $11.37 | 2.00× |
Each price comes from the provider's own Seedance 2.5 page, as of November 14, 2026: Replicate, Segmind, laozhang.ai, kie.ai, PiAPI, WaveSpeedAI and fal.ai. WaveSpeedAI shows a "10% OFF" badge without stating its conditions, which would bring its 720p clip to $3.24. Atlas Cloud lists Seedance 2.5 "starting from $0.134/second" without naming a resolution, so its price can't be placed in a resolution column.
Mainland China accounts on Volcengine pay ¥6.73, ¥15.12 or ¥37.42 for the same three clips (Volcengine pricing, as of November 14, 2026).
How the 10-second prices are calculated
BytePlus bills Seedance by video tokens, not by clip. Its formula is tokens = (input video seconds + output seconds) × width × height × frames per second ÷ 1,024. Seedance outputs 24 fps, and a 16:9 Seedance 2.5 clip is 854×480, 1280×720 or 1920×1080 pixels.
| Seedance 2.5, 10 s, 16:9 | Tokens | BytePlus rate per 1M tokens | Cost |
|---|---|---|---|
| 480p | 10 × 854 × 480 × 24 ÷ 1,024 = 96,075 | $10.70 | $1.03 |
| 720p | 10 × 1280 × 720 × 24 ÷ 1,024 = 216,000 | $10.70 | $2.31 |
| 1080p | 10 × 1920 × 1080 × 24 ÷ 1,024 = 486,000 | $11.7 | $5.69 |
The arithmetic matches BytePlus's own worked examples: a 5-second 720p Seedance 2.5 clip is listed at $1.156, or $0.231 per second. For resellers that price per second, the 10-second price is that rate × 10. laozhang.ai and fal.ai price per token, so their rates apply to the same token counts.
The table is a list-price estimate for one generation, not a bill. The amount you actually pay is set by usage.completion_tokens in the finished task, and counting 241 frames instead of 240 adds 0.42%. Input video, a model-chosen duration and discarded takes move the total much further, as covered in the cost section below.
BytePlus Seedance API pricing (official, per 1M tokens)
This is BytePlus ModelArk's official online-inference price list for Seedance as of November 14, 2026 (the page was last updated November 13, 2026).
| Model | Model ID | No input video | With input video | 10 s at 720p, no input video |
|---|---|---|---|---|
| Seedance 2.5 | dreamina-seedance-2-5-260628 | $10.70 (480p/720p), $11.7 (1080p) | $6.40 / $7.0 | $2.31 |
| Seedance 2.5 Flash (from Nov 15) | dreamina-seedance-2-5-flash-261115 | $7.0 (480p/720p only) | $4.3 | $1.51 |
| Seedance 2.0 | dreamina-seedance-2-0-260128 | $7.0 (480p/720p), $7.7 (1080p), $4.0 (4K) | $4.3 / $4.7 / $2.4 | $1.51 |
| Seedance 2.0 Fast | dreamina-seedance-2-0-fast-260128 | $5.6 (480p/720p only) | $3.3 | $1.21 |
| Seedance 2.0 Mini | dreamina-seedance-2-0-mini-260615 | $3.5 (480p/720p only) | $2.1 | $0.76 |
Seedance 2.5 has no 4K option. Seedance 1.0 Pro ($2.5 per 1M tokens) and 1.0 Pro Fast ($1 per 1M) are still listed, while Seedance 1.5 Pro no longer has a price row. If you came looking for the old 1.5 Pro price, the current low-cost official options are 2.0 Fast and 2.0 Mini.
What Is Seedance 2.0 and Why Its API Matters
Seedance 2.0 is now one model in ByteDance's Seedance API family: Seedance 2.5 is the newest and most expensive, Seedance 2.5 Flash arrives on November 15, 2026, and Seedance 2.0, 2.0 Fast and 2.0 Mini sit below it. "Dreamina Seedance" is the name BytePlus uses for the same models.
| Model | Resolutions | Length per request | 720p per second, BytePlus list |
|---|---|---|---|
| Seedance 2.5 | 480p, 720p, 1080p | 4–30 s (default -1: the model picks) | $0.231 |
| Seedance 2.5 Flash | 480p, 720p | not yet documented | $0.15 |
| Seedance 2.0 | 480p, 720p, 1080p, 4K | 4–15 s (default 5) | $0.15 |
| Seedance 2.0 Fast | 480p, 720p | 4–15 s (default 5) | $0.12 |
| Seedance 2.0 Mini | 480p, 720p | 4–15 s (default 5) | $0.08 |
Lengths and resolutions come from the BytePlus video generation API reference, and per-second prices from BytePlus's own 5-second examples. Seedance 2.5 and the 2.0 models generate audio by default (generate_audio: true), and audio carries no extra charge on the 2.x models. Which model fits which kind of shot is a separate decision, covered in Seedance Video Models in 2026: When to Use 2.5, 2.0, Fast, or Mini.
The API matters because it is the only way to price Seedance per clip and run it from your own code. Consumer apps such as Dreamina and Higgsfield sell Seedance through subscription credits, so their cost per video depends on the plan you pick and how much of it you use, and their credit tables can't be compared directly with a per-token API price. Through the API you submit an asynchronous task, poll it until it finishes and download the file, which is what makes batch production, retries and automated pipelines possible.
Official API vs Third-Party: Which Path Should You Choose?
The official BytePlus ModelArk API now sells Seedance 2.5 and the whole 2.0 family, so the old reason to use a reseller, getting a model before BytePlus offered it, is gone. On Seedance 2.5, BytePlus is never more expensive than any reseller that publishes per-resolution prices. A third party makes sense only when it saves you a step you can't or don't want to take, or when it prices a specific model lower.
| Your situation | Cheapest sensible route | What it costs you |
|---|---|---|
| You can prepay USD 30 and need Seedance 2.5 at any resolution, or in volume | BytePlus ModelArk | List price; credits can't be withdrawn |
| You already use Replicate and need 2.5 at 480p or 720p | Replicate | Same price as BytePlus; no 1080p |
| You need the lowest possible price per clip and 480p/720p is enough | Seedance 2.0 Mini on kie.ai | $0.19 (480p) or $0.41 (720p) per 10 s |
| You don't want a BytePlus account, or want Seedance billed alongside other models | laozhang.ai | About 12% above BytePlus on 2.5; failed or expired tasks refunded |
| You have a mainland China account | Volcengine Ark (CNY) | ¥200 activation and real-name verification |
laozhang.ai prices Seedance at Volcengine's CNY list price × 0.1714, which works out to about 1.2 times the Chinese list and $12.00 per 1M tokens for Seedance 2.5 at 480p/720p (laozhang.ai Seedance docs, as of November 14, 2026). That is $2.59 for a 10-second 720p clip, $0.28 more than BytePlus. It offers Seedance 2.5, 2.0, 2.0 Fast and 2.0 Mini, not yet 2.5 Flash, and it needs an API token created in the SeeDance2 group. The premium buys you a way around the BytePlus account and its USD 30 activation, one balance for Seedance and the other models on the same account, and an automatic refund of the pre-charge when a task fails or expires. If none of those matter to you, BytePlus is cheaper.
Enterprise contracts and SLAs are a separate question. BytePlus's Seedance docs publish no price for an enterprise trial or an SLA. What an enterprise-verified account gets by default is higher limits (600 requests per minute and 10 concurrent tasks instead of 180 and 3) and the option to buy Advanced Creation Rights, covered in the next section.
Seedance API Access: USD 30 to Activate, No $1 Million Deposit
To call Seedance 2.5, 2.0, 2.0 Fast or 2.0 Mini on BytePlus, you need any one of these on your account: at least USD 30 in Credits, a dedicated AI Savings Plan of at least USD 30, or Seedance resource packs worth at least USD 30 (BytePlus activation rules). The activation document does not require enterprise verification. Credits can't be withdrawn once bought, and Savings Plans and resource packs can't be canceled.
Resource packs are a way to meet the threshold, not a discount on Seedance 2.5. A 5M-token pack costs $32, but a 480p/720p Seedance 2.5 clip with no input video deducts 1.671875 pack tokens per video token (BytePlus resource pack rules). The pack therefore covers about 2.99M video tokens, which is the same $10.70 per 1M as paying as you go. Packs expire after 90 days and are non-refundable.
Do I need a $1 million deposit to get enterprise API access to Seedance?
No. No BytePlus or Volcengine page mentions a $1 million deposit, and Atlas Cloud's FAQ, updated August 25, 2026, calls the figure a myth rather than a pricing tier (Atlas Cloud). The official entry point is the USD 30 activation above.
The real enterprise-only purchase on BytePlus is Advanced Creation Rights: $14,000 a year or $1,400 a month, or $42,000 a year for the Premium tier (BytePlus purchase guide). It requires enterprise verification and buys larger private asset libraries for real-person and virtual-portrait workflows plus higher asset-upload rate limits. You don't need it to generate videos from text or images. An enterprise-only discount on Seedance 2.0 Fast and Mini ran from August 7 to October 7, 2026, and has ended.
Paying for Seedance without Alipay or WeChat Pay
You can. BytePlus bills in US dollars through a billing account set to either automatic billing with a bound payment method, or pay-by-credits, where you buy credits online from USD 5 or by bank transfer (BytePlus billing account). The docs don't list which card brands are accepted or which countries can open an account, so check both at signup before you count on the official route.
Volcengine, ByteDance's mainland China cloud, charges in CNY, needs a ¥200 balance, Savings Plan or resource pack to activate Seedance, and requires real-name verification for paid use (Volcengine activation). Unlike BytePlus packs, Volcengine Seedance packs can be refunded within 7 days if unused. On a reseller, what you can pay with is whatever its own checkout accepts.
Every Seedance API Provider Compared (Real Pricing Data)

For Seedance 2.5, no reseller with a per-resolution price list undercuts BytePlus. For the 2.0 family the order changes by model: kie.ai is cheapest for 2.0 Mini, PiAPI for 2.0 Fast, and BytePlus for Seedance 2.0 itself. The per-second prices below are each provider's own list as of November 14, 2026, for output with no input video.
| Provider | Seedance 2.5 (480p / 720p / 1080p) | Seedance 2.0 (480p / 720p / 1080p) | 2.0 Fast (480p / 720p) | 2.0 Mini (480p / 720p) |
|---|---|---|---|---|
| BytePlus (official) | $0.103 / $0.231 / $0.569 | $0.07 / $0.15 / $0.37 | $0.06 / $0.12 | $0.04 / $0.08 |
| Replicate | $0.1028 / $0.2312 / — | $0.08 / $0.18 / $0.45 | $0.07 / $0.15 | $0.04 / $0.09 |
| Segmind | $0.1082 / $0.2433 / $0.5985 | — | — | — |
| kie.ai | $0.140 / $0.315 / $0.790 | $0.095 / $0.205 / $0.51 | $0.059 / $0.124 | $0.019 / $0.041 |
| PiAPI | $0.15 / $0.35 / $0.80 | $0.10 / $0.20 / $0.50 | $0.048 / $0.096 | — |
| WaveSpeedAI | $0.18 / $0.36 / $0.90 | $0.12 / $0.24 / $0.60 | $0.10 / $0.20 | $0.06 / $0.12 |
BytePlus figures are the per-second rates from its own 5-second examples. fal.ai prices per token: $0.0214 per 1K tokens for 2.5 and $0.014 for 2.0 at 480p/720p, exactly twice BytePlus's rate. laozhang.ai's per-token prices run about 1.12–1.13 times BytePlus. Converted to one 10-second clip, the 2.0 family looks like this:
| 10-second clip, no input video | BytePlus | Replicate | kie.ai | PiAPI | laozhang.ai |
|---|---|---|---|---|---|
| Seedance 2.0, 720p | $1.51 | $1.80 | $2.05 | $2.00 | $1.70 |
| Seedance 2.0, 1080p | $3.74 | $4.50 | $5.10 | $5.00 | $4.25 |
| Seedance 2.0 Fast, 720p | $1.21 | $1.50 | $1.24 | $0.96 | $1.37 |
| Seedance 2.0 Mini, 480p | $0.35 | $0.40 | $0.19 | — | $0.40 |
| Seedance 2.0 Mini, 720p | $0.76 | $0.90 | $0.41 | — | $0.85 |
kie.ai Seedance pricing: kie.ai sells credits at $0.005 each and charges 63 credits per second for Seedance 2.5 at 720p, which is $0.315 per second, or 36% above BytePlus. Its 2.0 Mini rate of 3.8 credits per second at 480p is the lowest Seedance price in the tables above. You don't need kie.ai to reach Seedance through an API: BytePlus, Replicate and the other providers above all sell it directly.
PiAPI Seedance pricing: PiAPI charges $0.20 per second for Seedance 2.0 at 720p and $0.35 for 2.5 at 720p. A "Less Restriction" variant of 2.5 costs 10% more ($0.385 per second at 720p), and input video is billed at half the output rate.
Before you compare any other "from $x per second" price, check four things:
- Same model. A "Seedance" price may be 2.0 Mini, Fast or 2.5. Atlas Cloud's "from" prices ($0.134 per second for 2.5, $0.09 for 2.0, $0.045 for Mini) state no resolution, and Atlas listed 2.5 Flash at $0.109 per second before BytePlus opened its API.
- Same resolution. Ask which resolution a "from" price is for. A 720p Seedance 2.5 clip costs about 2.25 times a 480p clip, and a 1080p clip about 5.5 times.
- Same input rule. With a reference video, many providers bill the input seconds too. WaveSpeedAI charges $0.22 per second of reference plus output video at 720p.
- A model the upstream actually offers. WaveSpeedAI lists 4K for Seedance 2.5, which BytePlus says 2.5 does not support. Treat options that BytePlus doesn't sell as unconfirmed until you've generated one.
Ads and listings show other providers too, such as Protoface's "from $0.08/s". Treat any such price as unconfirmed until the provider shows which model, resolution and input rule it covers.
Stability and Reliability: The Metrics That Actually Matter
For Seedance 2.5 and the 2.0 family, the metrics that actually matter are the ones you can check: what you pay when a task fails, how many tasks you can run at once, and how long a task can wait. Uptime and success-rate figures are not among them, because BytePlus's Seedance docs publish none and neither do the reseller pricing pages listed above.

Failure billing. On BytePlus and Volcengine you are charged only for successfully generated videos, and tasks blocked by content moderation are free (BytePlus pricing). On laozhang.ai a request rejected with a 4xx error and no task ID isn't charged, and a task that ends as failed or expired has its pre-charge refunded at settlement. For other resellers, look for the failure rule on their billing page before you scale up, because a provider that charges for failed tasks raises your real cost per clip.
Concurrency and rate limits. On BytePlus, individual accounts get 180 requests per minute and 3 concurrent tasks for Seedance 2.5, 2.0, Fast and Mini by default, while enterprise accounts get 600 and 10. Seedance 2.0 at 4K is limited to 15 requests per minute and 1 concurrent task (BytePlus model list). With 3 concurrent tasks, a batch of 30 clips runs as 10 rounds, so queue your jobs rather than firing them all at once.
Task lifecycle. A Seedance task moves through queued, running, then succeeded, failed or expired. By default a task expires if it hasn't finished within 172,800 seconds (48 hours), set by execution_expires_after. Your polling code should treat failed and expired as final and record the error field.
Your own test. Before you commit a budget, submit 20–50 identical requests through the provider you're considering, spread over the day, and log each task's final status, time to finish and usage.completion_tokens. That gives you a success rate, a speed and a real cost per clip for your prompts, which no published table can.
Cost Optimization: Get More Videos for Less
The cheapest Seedance clip is the one you don't regenerate, so most savings come from controlling length, resolution and the number of discarded takes, not from switching providers. These are the things that make your bill differ from the 10-second table:
| Bill changer | What happens | Example, Seedance 2.5 on BytePlus |
|---|---|---|
duration: -1 (the 2.5 default) | The model picks a length between 4 and 30 seconds | A 30-second 720p clip is 648,000 tokens, or $6.93; set duration explicitly |
| Input video | Lower per-token rate, but input seconds are added and a minimum charge applies | 5 s of 720p output: $1.156 without video, $1.244–$4.838 with 2–30 s of input video |
| 1080p | 2.25 times the pixels plus a higher rate ($11.7 vs $10.70) | 10 s costs $5.69 instead of $2.31 (2.46×) |
| Discarded takes | Every successful generation is billed, used or not | Keeping 1 take in 3: $6.93 per usable 720p clip ($9.45 on kie.ai) |
| Turning audio off | No saving: audio has no surcharge on the 2.x models | Same price with generate_audio: false |
The 1-in-3 keep rate is an example, not a measured figure. Your own rate depends on the prompt and how strict you are, and it moves your cost per usable clip more than any price difference in the provider table.
Draft mode helps when you need a 1080p final. Seedance 2.5 can generate a 480p Draft (draft=true), billed like a normal 480p task, and then turn the Draft you pick into a final video, which can only be 1080p (BytePlus Seedance 2.5 guide). Draft task IDs stay valid for 14 days. As an example, three 10-second Drafts ($3.08) plus one 1080p final ($5.69) cost $8.77, against $17.06 for three blind 1080p takes. If 720p is enough, three direct 720p takes ($6.93) are cheaper than Draft plus a 1080p final.
Use the cheaper model for exploration. Testing prompt ideas on Seedance 2.0 Mini ($0.76 per 10-second 720p clip on BytePlus) or 2.0 Fast ($1.21) and moving to 2.5 only for finals lowers the cost of trial and error. The prompt won't carry over exactly, because the models behave differently. Seedance 2.5 Flash ($1.51 at 720p) will be another option from November 15, 2026, but its quality and speed against 2.5 are not documented yet.
Prepayment doesn't lower the 2.5 rate. BytePlus resource packs work out to the same $10.70 per 1M tokens as pay-as-you-go, as shown in the access section. On a reseller, compare its current price page before counting on any volume discount.
Quick Start Guide: Your First Seedance Video in 5 Minutes
You need a BytePlus API key with Seedance activated (USD 30 threshold), or a laozhang.ai token in the SeeDance2 group. A Seedance call is an asynchronous task: create it, poll it until it reaches a final status, then download content.video_url. The request shape below is BytePlus's own (BytePlus video generation API reference).
cURL: create a 10-second 720p Seedance 2.5 task
curl https://ark.ap-southeast.bytepluses.com/api/v3/contents/generations/tasks \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $ARK_API_KEY" \
-d '{
"model": "dreamina-seedance-2-5-260628",
"content": [
{"type": "text", "text": "Ocean waves crashing on a rocky coastline at sunset, aerial drone shot"}
],
"ratio": "16:9",
"duration": 10,
"resolution": "720p",
"generate_audio": true
}'
# The response contains the task id, for example {"id": "cgt-..."}
curl https://ark.ap-southeast.bytepluses.com/api/v3/contents/generations/tasks/<TASK_ID> \
-H "Authorization: Bearer $ARK_API_KEY"Python: create, poll, download and price the clip
This script needs requests (pip install requests) and an ARK_API_KEY environment variable. It prints the tokens actually billed and their list-price cost.
import os
import time
import requests
BASE_URL = "https://ark.ap-southeast.bytepluses.com/api/v3"
API_KEY = os.environ["ARK_API_KEY"]
MODEL = "dreamina-seedance-2-5-260628"
PRICE_PER_1M_TOKENS = 10.70 # Seedance 2.5, 480p/720p, no input video (BytePlus, Nov 14, 2026)
HEADERS = {"Authorization": f"Bearer {API_KEY}", "Content-Type": "application/json"}
body = {
"model": MODEL,
"content": [
{
"type": "text",
"text": "A golden retriever running through a sunlit meadow, "
"slow motion, shallow depth of field",
}
],
"ratio": "16:9",
"duration": 10, # set it: Seedance 2.5 defaults to -1 (model picks 4-30 s)
"resolution": "720p",
"generate_audio": True,
}
created = requests.post(f"{BASE_URL}/contents/generations/tasks",
headers=HEADERS, json=body, timeout=60)
created.raise_for_status()
task_id = created.json()["id"]
print("Task created:", task_id)
while True:
task = requests.get(f"{BASE_URL}/contents/generations/tasks/{task_id}",
headers=HEADERS, timeout=60).json()
status = task["status"]
if status in ("succeeded", "completed"):
break
if status in ("failed", "expired"):
raise SystemExit(f"Task {task_id} ended as {status}: {task.get('error')}")
print("Status:", status)
time.sleep(15)
tokens = task["usage"]["completion_tokens"]
print(f"Billed tokens: {tokens:,} -> ${tokens * PRICE_PER_1M_TOKENS / 1_000_000:.3f} at list price")
video_url = task["content"]["video_url"]
with open("seedance.mp4", "wb") as f:
f.write(requests.get(video_url, timeout=300).content)
print("Saved seedance.mp4")Running the same code on laozhang.ai: set BASE_URL to https://api.laozhang.ai/seedance/api/v3, use your SeeDance2-group token as the key, and change MODEL to doubao-seedance-2-5-260628; at laozhang.ai's $12.00 per 1M tokens, also set PRICE_PER_1M_TOKENS = 12.00. The request body is the same. laozhang.ai can return completed instead of succeeded, which the loop already accepts; if a finished task has no content.video_url, look for result_url or data.content.video_url (laozhang.ai Seedance docs).
Image-to-video uses the same task: add an item such as {"type": "image_url", "image_url": {"url": "https://..."}} to content, and write the text prompt as a motion instruction ("the character turns and smiles at the camera") rather than a description of what the image already shows.
Handling failures: a failed status from content moderation costs nothing on BytePlus; log the prompt and revise it rather than retrying it unchanged. For capacity problems, keep no more than 3 tasks running on an individual BytePlus account and queue the rest. For authentication, other endpoints and more parameters, see Seedance 2.0 API: Complete Developer Integration Guide (2026).
Seedance 2.0 vs Sora 2 vs Kling 3.0: Which Video AI Is Right for You?
Sora 2 is no longer an option: OpenAI closed the Sora app on April 26, 2026, and removed the Sora 2 API on September 24, 2026. The live comparison for Seedance is now against Google's Veo 3.1, Gemini Omni and Kling. On list price per 10-second 720p clip, Seedance 2.5 costs more than Kling 4.0, Veo 3.1 Fast and Gemini Omni 1.2 Flash, while Seedance 2.0 Mini is among the cheapest.
| Model and API | Price basis | 10 s at 720p | Price as of |
|---|---|---|---|
| Seedance 2.5 (BytePlus) | $10.70 per 1M tokens | $2.31 | November 14, 2026 |
| Seedance 2.0 (BytePlus) | $7.0 per 1M tokens | $1.51 | November 14, 2026 |
| Seedance 2.0 Mini (BytePlus) | $3.5 per 1M tokens | $0.76 | November 14, 2026 |
| Kling 4.0 (Kling API) | $0.157 per second | $1.57 | November 6, 2026 |
| Veo 3.1 Standard (Google Cloud, with audio) | $0.40 per second | $4.00 | November 5, 2026 |
| Veo 3.1 Fast (Google Cloud, with audio) | $0.10 per second | $1.00 | November 5, 2026 |
| Veo 3.1 Lite (Google Cloud, with audio) | $0.05 per second | $0.50 | November 5, 2026 |
| Gemini Omni 1.2 Flash (Gemini API) | $12.00 per 1M video output tokens, about $0.0695 per second | $0.70 | November 12, 2026 |
| Sora 2 | API removed September 24, 2026 | — | — |
Price says nothing about which model handles your shots best, and no shared quality score covers all of them. Kling's API now lists Kling 4.0 at $0.157 per second for 720p and $0.235 for 1080p, the same on fal.ai, Replicate and WaveSpeedAI. If you're deciding which model to test first, two side-by-side guides go deeper: Seedance 2.5 vs Gemini Omni: Which Should You Test First? and Seedance 2.5 or Kling 3.0: Which Should You Test First?. If you still have a Sora 2 workflow to replace, start with Cheapest Sora 2 API After the Shutdown: What to Use Instead.
A practical split for most teams: draft and iterate on the cheaper models, and spend Seedance 2.5's $2.31 per 10-second 720p clip only on shots where you need its 30-second length or its output.
Frequently Asked Questions
Is Seedance 2.0 API officially available yet?
Yes. BytePlus ModelArk sells Seedance 2.0 (dreamina-seedance-2-0-260128), 2.0 Fast, 2.0 Mini and Seedance 2.5 through its official API, with Seedance 2.5 Flash callable from 14:00 UTC+8 on November 15, 2026. You activate them by holding USD 30 in credits, a Savings Plan or a resource pack. In mainland China, Volcengine offers the same models under doubao-seedance-… IDs.
How much does it cost to generate a Seedance video?
As of November 14, 2026, a 10-second 16:9 Seedance 2.5 video costs $1.03 at 480p, $2.31 at 720p and $5.69 at 1080p at BytePlus list price, and up to twice that on resellers. Seedance 2.0 costs $1.51 for the same 720p clip, and 2.0 Mini $0.76 ($0.41 on kie.ai). These figures assume no input video and one generation; discarded takes and model-chosen lengths raise the real cost.
How much is Seedance 2.5 per video?
It depends on length and resolution, because Seedance 2.5 is billed per token, not per video. A 5-second 720p clip is $1.156 on BytePlus, a 10-second clip $2.31, and the 30-second maximum $6.93. At 1080p, a 10-second clip is $5.69.
Why is Seedance 2.5 so expensive?
Two things add up. Seedance 2.5's rate of $10.70 per 1M tokens is 53% above Seedance 2.0's $7.0, and tokens grow with every pixel and frame: a 10-second 720p clip at 24 fps is 216,000 tokens. The result, $0.231 per second at 720p, is more than double Veo 3.1 Fast ($0.10) and above Kling 4.0 ($0.157). Buying through a reseller adds up to 100% on top, and every discarded take is billed in full.
Is the Seedance API free, and is there a free Seedance 2.5 API key?
Not in any documented amount. BytePlus gives new users a free inference quota per model and has a "Free Tokens Only" mode capped at 500,000 free tokens (BytePlus free quota), but no BytePlus document states a free amount for Seedance 2.x, and paid Seedance use starts with the USD 30 activation. For scale, one 10-second 720p clip uses 216,000 tokens. A "2M free tokens" figure that circulates for Seedance is not documented for the 2.x models.
What is the success rate for Seedance video generation?
No official figure exists: BytePlus's Seedance docs publish no success rate or uptime, and the reseller pricing pages listed above don't either. What is documented is that BytePlus and Volcengine charge only for successfully generated videos and don't charge for moderation failures, and that laozhang.ai refunds failed or expired tasks. To get a success rate that means something for your prompts, run 20–50 test tasks and log their final status.
Can I use Seedance 2.0 for commercial projects?
BytePlus sells Seedance as a paid commercial API, and its pricing and activation docs don't add a separate commercial-use fee for generated videos. Your rights to the output are set by the BytePlus service terms, or by your reseller's terms if you buy through one, so read those before a client project. Videos of real people are handled separately: BytePlus requires real-person verification to upload real-person assets, and larger real-person and portrait libraries are part of the paid Advanced Creation Rights.
How does Seedance compare to Sora 2 in quality?
That comparison can no longer be run on the API: OpenAI removed the Sora 2 API on September 24, 2026. Any Sora 2 score you find predates the shutdown and can't be reproduced. The current alternatives are Veo 3.1, Gemini Omni and Kling 4.0; for price, see the comparison table above, and for a first test, Seedance 2.5 vs Gemini Omni: Which Should You Test First?.
How long of a video can Seedance 2.5 make?
Seedance 2.5 accepts 4 to 30 seconds per request. If you don't set duration, it defaults to -1 and the model chooses the length, which also decides the price. The Seedance 2.0 models accept 4 to 15 seconds and default to 5.






