How to Choose a Seedance 2.5 API: Providers, Pricing, and Integration Compared
Compare published Seedance 2.5 API prices from Kinovi, BytePlus, Replicate, Pika, WaveSpeed and fal at 480p through 4K, then check Kinovi's credit costs, submit and recordInfo flow, and what each provider's integration actually requires.

On this page
Choosing Seedance 2.5 is only the first step. On platforms such as Kinovi, WaveSpeed, fal, Replicate, Pika, OpenRouter, Segmind, APIMart, and Kie, developers care more about what an API call actually costs for the same video, how much existing code has to change, and how requesting a key, submitting a job, uploading assets, and polling or receiving a callback differ from one provider to the next. This article focuses on Kinovi and compares each provider's published API prices and integration flow, so you can work out the cost and see what integration involves.
Prices come from each provider's API docs or model page. The comparison uses API list prices only, not web membership or subscription rates. Seedance 2.5 Flash is a lower-priced model in the same family and is not mixed with 2.5.
In this article
Start with a same-condition cost comparison, then see what it costs to generate a video on Kinovi and how to call the API. Adding a reference video often changes the unit price and the billable duration, so a "lowest starting price" is not enough on its own. fal bills by token, and the per-second figures in its docs are only rough estimates.
1. Start with a provider price comparison
The table below lists published unit prices for 480p, 720p, 1080p, and 4K. Kinovi is shown first; the other providers are sorted by the USD cost of a 5-second, 720p, 16:9 video with no reference video. Provider names link to the model page, with Pricing and API documentation links beside them.
| Provider | 480p | 720p | 1080p | 4K | 5s 720p | Cost vs. Kinovi | Billing |
|---|---|---|---|---|---|---|---|
| Kinovi Pricing · API | $0.1197/s (26 credits/s) | $0.2717/s (59 credits/s) | $0.6604/s (143.4 credits/s) | $1.0964/s (238.08 credits/s) | 295 credits · $1.36 | Baseline | Credits per second; the model page also publishes the list price |
| OpenRouter Pricing · API | $0.1028/s | $0.23112/s | Not listed separately | ~$2.08/s | $1.1556 | 14.9% less | Billed by token; $10.70 / 1M tokens without video input, $6.40 / 1M with video input; 720p rate is from the provider table; 4K is from the cost FAQ |
| BytePlus ModelArk (official) Pricing | $0.1028/s | $0.2312/s | $0.5686/s | $0.9136/s | $1.156 | 14.9% less | Billed by token; table figures are official 16:9 five-second examples |
| Replicate Pricing · API | $0.1028/s | $0.2312/s | Not listed | Not listed | $1.156 | 14.9% less | Per output second |
| Pika API Pricing · API | ~$0.1085/s | ~$0.2427/s | ~$0.5971/s | ~$1.005/s | ~$1.2135 | 10.7% less | Billed by token; per-second figures in the table are derived estimates; 480p/720p is $11.235 / 1M tokens, 1080p is $12.285 / 1M tokens, and 4K is $5.17 / 1M tokens |
| Segmind Pricing · API | $0.1082/s | $0.2433/s | $0.5985/s | Not listed | ~$1.2266 | 9.7% less | Billed by token ($11.263141 / 1M at 720p with no video input); 16:9 per-second rates are from the pricing page; the 5s total includes the one extra billed frame |
| Higgsfield Pricing · API | $0.1177/s | $0.2647/s | $0.6518/s | Not listed | $1.3235 | 2.6% less | Per output second (standard self-serve API list rate) |
| APIMart Pricing · API | $0.1201/s | $0.27/s | $0.6683/s | Not listed | $1.35 | 0.6% less | Official list rates, not Gold or other membership discounts; 480p/720p is $12.50 / 1M tokens without a reference video, and $7.50 / 1M with a reference video |
| Kie Pricing · API | $0.140/s (28 credits/s) | $0.315/s (63 credits/s) | $0.790/s (158 credits/s) | Not listed | $1.575 | 15.9% more | Credits per second; 1 credit = $0.005; list rates, not top-up bonuses |
| WaveSpeed Pricing · API | $0.18/s | $0.36/s | $0.90/s | $1.80/s | $1.80 | 32.5% more | Per output second (model-page list price) |
| fal Pricing · API | ~$0.2205/s (estimate) | ~$0.4730/s (estimate) | ~$1.164/s (estimate) | Not listed separately | ~$2.365 (estimate) | 74.1% more | Billed by token: 480p/720p is $0.0214 / 1,000 tokens, and 1080p is about $0.0234 / 1,000 tokens |
Also keep these billing conditions in mind:
- The table above covers text-to-video or image-to-video with no reference video. The next table prices the same 720p job with a reference video.
- Replicate's
video_intier is billed on output seconds only, at $0.4304/s for 480p and $0.9676/s for 720p, which is a different scheme from "a lower unit price that also includes input duration." - With a reference video, WaveSpeed lists 480p at $0.11, 720p at $0.22, 1080p at $0.55, and 4K at $1.10. Each reference clip is billed for at least 2 seconds, each clip and the combined total are capped at 30 seconds, and the combined duration is rounded up and billed with the output duration at the same reference-tier rate. The model page also shows 10% OFF; the table uses list prices.
- BytePlus bills by token and has lower unit prices and minimum-usage rules when video input is included.
- Pika API bills by token; the per-second and 5-second figures in the no-reference table are derived estimates. With a reference video, use the reference-to-video endpoint. Rates are $6.72 / 1M tokens for 480p/720p, $7.35 / 1M tokens for 1080p, and $3.19 / 1M tokens for 4K, and input duration is also billed. Only successful generations are charged, and pricing includes a 5% API Club platform fee.
- OpenRouter bills by token at $10.70 / 1M without video input and $6.40 / 1M with video input. The 720p no-reference total uses the published $0.23112/s provider-table rate. The cost FAQ also lists 4K at $2.08/s; 1080p is not listed separately.
- Segmind bills by token and publishes 16:9 per-second rates. A 5-second 720p clip is billed for one extra frame on top of those five seconds. With a reference video, billed seconds are output duration plus the total reference duration, with a minimum reference portion of two-thirds of the output duration, rounded up. Failed jobs are not billed.
- APIMart figures use the Official list rates, not Gold, Platinum, or Diamond discounts. 4K is not listed. With a reference video, input duration is billed with the output at the reference token rate.
- Kie bills credits per second (1 credit = $0.005). With video input, the rate is multiplied by uploaded video duration plus generated duration. 4K is not listed. Top-up bonuses are not used in the table.
The table below prices a 720p, 16:9 job with a 10-second reference video and 5-second output. Kinovi is shown first; the other providers are sorted by that total.
| Provider | 720p rate used | 10s ref + 5s out | Cost vs. Kinovi | Billing |
|---|---|---|---|---|
| Kinovi Pricing · Docs | $0.1612/s (35 credits/s) | 525 credits · $2.418 | Baseline | (10 + 5) × reference-tier rate |
| Higgsfield Pricing · API | $0.3264/s | $1.632 | 32.5% less | Output seconds only at the with-video-input list rate; the 10s reference is not multiplied into the total |
| BytePlus ModelArk (official) Pricing | $6.40 / 1M tokens | ~$2.0736 | 14.2% less | Official token formula on (10 + 5) seconds; 10s input is above the 2–4s minimum |
| OpenRouter Pricing · API | $6.40 / 1M tokens | ~$2.0736 | 14.2% less | Official with-video token rate on (10 + 5) seconds at 1280×720, 24 fps |
| Pika API Pricing · API | $6.72 / 1M tokens | ~$2.1773 | 10.0% less | Derived estimate on the reference-to-video endpoint; input duration is billed |
| Segmind Pricing · API | $6.736832 / 1M tokens | ~$2.1827 | 9.7% less | (10 + 5) seconds at the video-input token rate; 10s is above the two-thirds minimum |
| APIMart Pricing · API | $7.50 / 1M tokens | $2.43 | 0.5% more | Official list rate on (10 + 5) seconds; membership discounts are not used |
| Kie Pricing · API | $0.190/s (38 credits/s) | $2.85 | 17.9% more | (10 + 5) × with-video-input list rate |
| WaveSpeed Pricing · API | $0.22/s | $3.30 | 36.5% more | (normalized 10 + 5) × list rate; promo prices are not mixed in |
| fal Pricing · API | $0.0214 / 1,000 tokens (× 0.6) | ~$4.1602 | 72.0% more | Reference-to-video endpoint; billed on input plus output duration |
| Replicate Pricing · API | $0.9676/s (video_in) | $4.838 | 100.1% more | Output seconds only; the 10s reference is not multiplied into the total |
Token-priced totals are estimated at 1280×720, 24 fps.
2. What does it cost to generate a video on Kinovi?
Kinovi's model ID is seedance2-5. Fees depend on output resolution and whether a reference video is included, and are billed in credits. The model page pricing section also publishes the list price.
| Tier | Credits per second | Published list price | 5-second cost (no reference video) |
|---|---|---|---|
| 480p | 26 | $0.1197/s | 130 credits · $0.5985 |
| 480p + reference video | 16 | $0.0737/s | Settled as (reference video seconds, up to 30) + output duration |
| 720p | 59 | $0.2717/s | 295 credits · $1.36 |
| 720p + reference video | 35 | $0.1612/s | 10s reference + 5s output = 525 credits · $2.42 |
| 1080p | 143.4 | $0.6604/s | 717 credits · $3.302 |
| 1080p + reference video | 85.68 | $0.3947/s | Settled as (reference video seconds, up to 30) + output duration |
| 4K | 238.08 | $1.0964/s | 1190.4 credits · $5.482 |
| 4K + reference video | 145.92 | $0.6720/s | Settled as (reference video seconds, up to 30) + output duration |
The 720p 5-second and reference-video examples are in the Seedance 2.5 docs. With no reference video, you are billed at submit time as output duration × the matching unit price. With a reference video, a reservation is taken at submit based on output duration, then settled after success as (reference video seconds, up to 30, + output duration) × the reference-tier unit price. Do not multiply output seconds alone by the reference-tier rate. Reference images and reference audio do not change the unit price.
According to Credits and Billing and API Reference: Studio and the API share the same credit balance, and purchased credits do not expire until they are used. A reservation is taken at submit for the selected settings; if the balance is insufficient, the request returns 402 and nothing is charged. After success, creditsUsed on recordInfo is authoritative. Failures, timeouts, and cancellations refund the reservation. Concurrency is limited by account tier; new accounts can run 3 jobs at once by default, and exceeding that returns 429.
If you already generate on Kinovi, or you want the console and the API to draw from the same balance, estimate the cost on the model page by resolution and whether a reference video is included, then call the API with the same inputs. For product shots, character clips, or ads with multiple reference assets, you can tune the prompt at 480p or 720p first, then raise the resolution to 1080p or 4K once the look is right.
3. How do you integrate each API?
Kinovi's model page API section and API Reference, along with the Seedance 2.5 docs, document the full flow:
- Create a key in API Keys. Send it as
Authorization: Bearer YOUR_API_KEY. - Upload local assets first with
POST /api/v1/uploadsto get an accessible URL (see Uploading Assets). Uploaded files are deleted after 24 hours, so upload them just before submit and do not cache that URL long term. Generated outputs are not subject to this expiry rule. POST https://kinovi.ai/api/v1/jobs/createTasksubmits the job, returns ataskIdimmediately, and reserves credits.- Poll
GET https://kinovi.ai/api/v1/jobs/recordInfo?taskId=untilstatusissuccessorfail. - You may optionally pass a top-level
callBackUrl. When the job reaches a terminal state, Kinovi POSTs once. The docs state that this callback is not retried, so you should still pollrecordInfoas the source of truth.
Minimal submit example (replace YOUR_API_KEY with your own key):
curl -X POST https://kinovi.ai/api/v1/jobs/createTask \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "seedance2-5",
"inputs": {
"prompt": "A golden retriever running through autumn leaves in slow motion, cinematic depth of field, warm afternoon light.",
"duration": 5,
"outputResolution": "720p",
"aspectRatio": "16:9"
}
}'
curl "https://kinovi.ai/api/v1/jobs/recordInfo?taskId=YOUR_TASK_ID" \
-H "Authorization: Bearer YOUR_API_KEY"
After success, take the video from output[].url. creditsUsed is the amount actually charged for the job.
Every provider uses an async job, but auth headers, paths, and field names are not interchangeable. You cannot migrate by changing only the endpoint.
| Platform | Auth | Submit | Poll / callback | Docs |
|---|---|---|---|---|
| Kinovi | Authorization: Bearer | POST /api/v1/jobs/createTask | GET /api/v1/jobs/recordInfo?taskId=; optional callBackUrl (sent once, not retried) | Model page API · Docs |
| BytePlus | Bearer | Async generations job, model ID dreamina-seedance-2-5-260628 | Poll by job ID | Pricing and docs |
| WaveSpeed | Authorization: Bearer | POST https://api.wavespeed.ai/api/v3/bytedance/seedance-2.5/text-to-video | Poll GET https://api.wavespeed.ai/api/v3/predictions/{id}/result | Submit and poll |
| fal | API Key | The docs example is fal.subscribe('bytedance/seedance-2.5/us/text-to-video', …) | The client reports status and results through the subscription | Submit a request |
| Replicate | Authorization: Bearer ($REPLICATE_API_TOKEN) | POST https://api.replicate.com/v1/models/bytedance/seedance-2.5/predictions | GET /v1/predictions/{id}; the docs support webhooks | API reference |
| Pika API | X-API-Key | POST https://api.dev.pika.art/v1/media/bytedance/seedance-2.5/text-to-video | Poll GET https://api.dev.pika.art/v1/media/jobs/{request_id} | Model page · Pricing |
| OpenRouter | Authorization: Bearer | POST https://openrouter.ai/api/v1/videos, model bytedance/seedance-2.5 | Poll the returned polling_url, or GET /api/v1/videos/{jobId} | Submit a request · Video generation |
| Segmind | API key (SEGMIND_API_KEY) | POST https://api.segmind.com/v2/seedance-2.5 | Poll GET /v2/requests/{id}/status until COMPLETED or FAILED | API · Pricing |
| APIMart | Authorization: Bearer | POST https://api.apimart.ai/v1/videos/generations, model seedance-2.5 | Poll GET https://api.apimart.ai/v1/tasks/{task_id} | Generation · Task status |
| Kie | Authorization: Bearer | POST https://api.kie.ai/api/v1/jobs/createTask | Poll GET https://api.kie.ai/api/v1/jobs/recordInfo?taskId=; optional callBackUrl | Docs · Model page |
BytePlus
WaveSpeed
fal
Replicate
Pika API
4. Where does Kinovi have an advantage?
On comparable unit prices for a 5-second 720p video with no reference video, Kinovi is not the cheapest option: Volcengine Ark converts to about $1.126, BytePlus and Replicate are about $1.156, and the Kinovi docs example is 295 credits, about $1.36. Its strengths are mainly in how clearly billing is explained and how complete the integration is:
- Studio and the API share the same balance; purchased credits do not expire until they are used, so the console and the API do not need separate top-ups.
- The model page lists both credits and the USD price by resolution, and the docs include settlement examples for 5 seconds and for "10 seconds of reference + 5 seconds of output," so a reference video is not undercounted by output seconds alone.
- Failures, timeouts, and cancellations refund the reserved credits; if the balance is insufficient, the submit is rejected and nothing is charged.
- Creating a key, uploading assets, submitting a job, polling for the result, and the optional callback are all publicly documented. The callback is sent once and is not retried; polling
recordInfois sufficient as the source of truth. - The model page already publishes 4K-tier prices, so you can estimate the budget against the final output spec.
5. How should developers choose?
You can choose based on project constraints, not only on which platform has the lowest list price:
- If you already generate on Kinovi, or you need Studio and the API to share a balance and want credits and list prices visible by resolution: use
seedance2-5and the endpoints above. - If you must use an official ByteDance account, need token-based billing, or have to meet a regional onboarding requirement: evaluate BytePlus, starting with the official onboarding docs.
- If you are comparing other aggregators: first fix resolution, duration, and whether a reference video is included, then check whether the other side bills by output second, token, or credit; also confirm whether failures are refunded, how assets are uploaded, and whether callbacks are retried. fal needs to be estimated by token.
Model page and API entry points: Seedance 2.5 · API docs · API Keys
FAQ
How much does a 5-second 720p API call cost on Kinovi?
With no reference video, the docs example is 295 credits, about $1.36. With a 10-second reference video, it is billed as 15 seconds × 35 credits, or 525 credits, about $2.42.
Can I migrate to another platform by changing only the API URL?
No. Auth headers, paths, and fields are not interchangeable across providers. You need to rewrite the request against the other provider's docs and re-price the job under their rules.
Can I use a website membership rate or credits to estimate API cost?
Not directly. The comparison uses each provider's API list prices only, not web membership or subscription rates.
Related articles
More to try
Related models

ByteDance · video model
Seedance 2.5
Seedance 2.5 is ByteDance’s next-generation AI video model for up to 30-second generation, multimodal references, and precise local editing.
$0.1197/s

ByteDance · video model
Seedance 2.0
Next-generation AI video model by ByteDance. Combines Fast and Pro modes for high-quality video generation.
$0.0707/s

NanoBanana · image model
NanoBanana 2
Fast, high-quality AI image generation with creative control and photorealistic output.
$0.0707/image


