Introduction
Seedance 2.0 can turn a prompt, an image, or a set of reference assets into a short video. For an AI agent, however, the model is only part of the workflow. The agent also needs a reliable way to submit a job, track its status, retrieve the finished file, and account for the cost. If the agent is connected through MCP, the API must fit into a tool workflow where one step can take minutes rather than milliseconds.
This guide compares ten services that provide access to Seedance 2.0 or Seedance 2.0 Fast. A faster model can make sense for drafts and high-volume variations; the standard model may be the better choice when the final output calls for higher resolution or tighter visual detail. The right provider depends on the input controls, task interface, billing rules, and the way your agent works. For a look at a specific model endpoint, see the Seedance 2.0 API listing.
The prices below describe the configurations in the supplied platform material. They are examples for comparing billing behavior, not a promise that every account, channel, or future request will receive the same quote. Check the live rate and supported parameters before building a production budget.
How We Evaluated These APIs
We reviewed the supplied API specifications and pricing notes against five practical questions an agent developer would ask:
- Agent workflow: Can a program submit a job, receive an ID, check progress, and collect a result? Webhooks and callbacks help when polling would waste requests.
- MCP fit: Does the service offer an explicit MCP route, or can its API be wrapped as an MCP tool with clear inputs and outputs? An MCP server is a separate integration layer; a REST endpoint alone does not mean native MCP support.
- Creative controls: We looked for text, first and last frames, reference images, video, audio, duration, aspect ratio, and audio generation. The available combination matters more than a long feature list.
- Cost clarity: We compared the stated unit price with a concrete five-second example where possible. Reference-video inputs, resolution, channel choice, membership tier, and credit conversion can change the charge.
- Operational fit: We considered model switching, failed-job billing, concurrency information, delivery format, and whether a team can estimate costs before an agent launches several jobs.
This is a documentation-based comparison, not a timed benchmark or hands-on reliability test. Reported generation times and provider claims are identified as such. The list is an editorial ordering for different use cases, not a universal speed or quality ranking.
TL;DR: Which Provider Fits Which Workflow?
|
Rank |
Provider |
Practical strength |
Illustrative price from supplied material |
MCP route |
|---|---|---|---|---|
|
1 |
ApiPass |
Channel selection and detailed reference controls |
Fast, 5s at 720p without video input: $0.025 on the listed Starter route |
REST API or SDK wrapper |
|
2 |
Monid |
Agent-oriented tool catalog |
Standard, about $0.76 for 5s at 720p |
Native MCP option |
|
3 |
Runware |
Broad video task and delivery options |
Fast baseline: $0.06 per second at 480p |
API wrapper |
|
4 |
OpenRouter |
One API pattern for model routing |
Fast, 5s at 480p: about $0.20 at listed rate |
API wrapper |
|
5 |
APIMart |
Reference-rich requests and returned last frame |
Standard, 5s at 720p: 7.10 credits at listed Gold rate |
API wrapper |
|
6 |
Replicate |
Simple per-output-second pricing |
Fast, 5s at 720p: $0.75 |
API wrapper |
|
7 |
OceanToken |
Native versus upscaled rendering choice |
Standard, 5s at 720p upscaled: $0.535 at Tier 1 |
API wrapper |
|
8 |
WaveSpeed |
Separate model endpoints and predictable rates |
Fast, 5s at 720p: $1.00 list price |
API wrapper |
|
9 |
Seedance API |
Fixed-duration jobs and subscription options |
Fast, 5s at 720p: $2.00 at pay-as-you-go rate |
API wrapper |
|
10 |
APIPod |
Multimodal modes and token-based accounting |
Fast from $0.045 per second in its stated reference rate |
API wrapper |
The prices in this table are intentionally tied to their stated settings. They do not make a like-for-like price league table: resolution, render mode, membership, and whether input video is billable differ across providers.
Website List
1. ApiPass
ApiPass is a managed API layer for the Seedance 2.0 family. Its asynchronous job API returns a task that an agent can monitor through status queries or a webhook callback. A JavaScript and TypeScript SDK is available alongside REST calls. The platform also exposes a channel setting: Starter for low cost, Regular for a steadier standard route, Official for the native route, and Auto for automatic selection. That lets a workflow choose a route according to the job’s budget and delivery needs.
- Features: Requests can include a prompt, first and last frames, up to nine reference images, three reference videos, and three reference audio clips within the documented size and duration limits. Prompts can bind assets with @image1, @video1, and @audio1. Output duration runs from 4 to 15 seconds, with aspect ratios including 16:9, 9:16, square, and adaptive. Synchronized audio and web search are configurable.
- Pricing: The supplied Starter listing gives Fast at $0.005 per output second without reference video for both 480p and 720p, making a five-second job $0.025. Standard at 480p is listed at $0.045 per second, or $0.225 for five seconds. A Regular 720p standard example is $0.236 per second, or $1.18 for five seconds. With reference video, the charge can include input and output duration.
- Advantages: The named channel choice, callbacks, SDK, and explicit input limits help an agent make a deliberate request and track the result. Asset tags are useful when a prompt refers to several shots or sounds.
- Limits: The first/last-frame mode and multimodal-reference mode are mutually exclusive in the supplied specification. Costs also depend on channel and reference-video duration, so the agent should calculate a quote for each configuration.
Best for: Teams that want detailed reference control and a route choice inside one managed job API. The Seedance 2.0 fast endpoint is especially relevant to prompt iteration and batches of short previews.
2. Monid
Monid presents itself as an agent tool platform, with one key for a large catalog that includes video generation. It is the clearest native MCP candidate in this list: the supplied material names Skill, MCP, and CLI as three ways to connect. That can reduce the integration work for an agent already using a remote MCP server, although a developer should still inspect the particular Seedance tool schema and its returned job details.
- Features: Its catalog is described as more than 2,900 tools from 72 providers, with 39 video endpoints. It lists Seedance standard, Fast, Mini, and a later audio-capable model alongside other tools an agent may call in one workflow.
- Pricing: Monid states no subscription and no added platform markup. Its listed Seedance 2.0 rate is $7 per million tokens, with an estimated $0.76 for a five-second 720p clip. Fast is $5.60 per million tokens, estimated at $0.60 for the same example. A $1 trial credit is listed.
- Advantages: Native MCP access and an upfront per-call quote suit agents that make decisions across multiple tools. Token pricing can also be visible before a call.
- Limits: The per-video figures are estimates based on token use. A workflow should use its quoted call cost rather than assume every five-second output uses exactly the sample number of tokens.
Best for: Agent builders who want Seedance inside an existing MCP or CLI tool collection.
3. Runware
Runware is a hosted inference API with a broad set of video task types. Its Seedance entries include text-to-video, image-to-video, video-to-video, audio-to-video, editing, extension, and native audio generation. That range is useful when one agent needs more than prompt-to-video output.
- Features: Requests may use first and last frames and image, video, or audio references. The supplied material describes configurable output dimensions, MP4, WEBM, or MOV delivery, synchronous and asynchronous modes, and webhook callbacks.
- Pricing: The stated 480p baseline is $0.070 per second for standard and $0.060 per second for Fast. The platform examples vary by task: a Fast eight-second first/last-frame job is about $0.4846, while a ten-second Fast multimodal-reference job is about $1.2884.
- Advantages: Multiple delivery modes and formats can save downstream conversion steps. A webhook can hand a finished asset back to an agent without continuous polling.
- Limits: A baseline per-second number does not fully describe jobs with different duration, resolution, frame rate, or reference complexity. Use the configuration-specific price shown for the intended request.
Best for: Production pipelines that mix generation, editing, extension, and different output formats.
4. OpenRouter
OpenRouter offers a unified REST pattern for video jobs. In the supplied specification, an agent submits a request to /api/v1/videos, receives a job ID and polling URL, then retrieves the completed file URL. That familiar request-and-poll sequence is straightforward to expose as an MCP tool, but the Seedance route described here is an API integration rather than a native MCP server.
- Features: The notes list prompt, first and last frame, reference video, and reference audio inputs, with 4 to 15 seconds of output. They also describe resolution tiers from 480p to 4K and native audio generation. Check model-specific support for the chosen tier before sending a job.
- Pricing: The listed 480p rates are $0.06726 per second for standard and $0.04035 for Fast: about $0.336 and $0.202 for five seconds. Another supplied estimate uses token rates of $7 and $4.20 per million tokens, respectively. The input-video rate follows a different schedule.
- Advantages: One submission and polling pattern simplifies the agent wrapper. Published per-second and token figures help estimate a budget across output sizes.
- Limits: Different pricing presentations in the source material produce different five-second examples. Tie a cost estimate to the exact model, resolution, and input type shown at checkout or in the API response.
Best for: Developers already routing model calls through OpenRouter who want video in the same integration style.
5. APIMart
APIMart exposes standard, Fast, and Mini through one API key. An agent submits a video generation job, polls a task endpoint, and receives the finished URL. Its request controls are detailed enough to support a sequence of related clips, not just isolated prompts.
- Features: The supplied specification lists up to nine images, three reference videos, three reference audio clips, first and last frame roles, and prompt binding. return_last_frame can make the end of one clip the starting asset for the next. Audio generation is on by default, and the duration range is 4 to 15 seconds.
- Pricing: APIMart uses credits and member tiers. Its sample Gold rate for standard 720p without video input is 1.42 credits per second, so a five-second output is 7.10 credits. Fast at 720p is 0.856 credits per second, or 4.28 credits for five seconds. Input-video jobs count both input and output duration under their separate rate.
- Advantages: The returned last frame is a practical aid for multi-shot agent workflows. The explicit media caps and aspect-ratio options make request validation possible before submission.
- Limits: The supplied credit-to-dollar examples are internally inconsistent, so credits are the safer comparison unit here. Membership level and input-video duration also affect the final debit.
Best for: Teams building connected shots and tracking usage in platform credits.
6. Replicate
Replicate hosts Seedance models with a direct per-second output price in the supplied notes. Its interface supports both basic text or image generation and richer multimodal references. For an agent, a short list of input fields and a predictable output-duration charge can make the tool definition easy to reason about.
- Features: It supports first frame, optional last frame, up to nine reference images, three videos, and three audio clips under the stated limits. Prompt references such as [Image1] and [Video1] identify assets. The notes describe native audio, editing, extension, and an intelligent-duration option.
- Pricing: Fast without video input is listed at $0.07 per second at 480p and $0.15 at 720p, or $0.35 and $0.75 for five seconds. Standard is $0.08 and $0.18 per second at those resolutions, or $0.40 and $0.90 for five seconds. Reference-video rates are separate.
- Advantages: Per-output-second figures make a first budget estimate easy. The documented reference syntax is useful when an agent composes a prompt from several uploaded assets.
- Limits: Some advanced controls differ by model. In the supplied material, Fast is limited to 480p and 720p, while standard has the higher-resolution options.
Best for: Developers who want a compact hosted model interface and clear output-duration examples.
7. OceanToken
OceanToken offers two rendering choices that should be treated as separate products in a quality review. Native passes the prompt and references to the original model route. Upscaled adds the provider’s super-resolution step after base rendering. The latter may suit assets where a larger delivery frame matters more than native fine detail.
- Features: The REST job flow returns an ID, supports status polling, and offers a content endpoint for the finished file. The notes list 4 to 15 seconds of output and up to nine reference images. A failed job releases its temporary hold instead of completing a charge.
- Pricing: Prepaid balance is valid for 12 months, with tiers based on cumulative top-ups. At Tier 1, standard 720p Upscaled is listed at $0.107 per second, or $0.535 for five seconds. Standard 720p Native is $0.144 per second, or $0.72 for five seconds. The supplied Fast example is 1080p Upscaled at $0.192 per second.
- Advantages: A named Native/Upscaled choice lets the agent select the rendering path for each job. Failed-generation billing is explicit.
- Limits: Upscaled resolution does not imply the same source detail as native generation at that size. Tier discounts and render mode must be included in any cost comparison.
Best for: Workflows that can choose between native detail and a lower-cost enlarged output on a shot-by-shot basis.
8. WaveSpeed
WaveSpeed provides separate REST endpoints for standard and Fast. An agent submits a prediction, receives an ID, polls for completed, and extracts the output URL. The material describes a full reference set and a duration range wide enough for short ad variants or storyboard clips.
- Features: Listed controls include audio generation, aspect ratio, resolution, 4-to-15-second duration, web search, and up to nine images, three videos, and three audio references. Tags such as @Image 1 can tie a prompt to a reference.
- Pricing: The supplied list price for Fast is $0.20 per second at 720p, or $1.00 for five seconds; standard is $0.24 per second, or $1.20. With reference video, the platform counts normalized input-video time plus output time at a different per-second rate.
- Advantages: The separate endpoint and prediction ID make it clear which model a tool is calling. A linear price schedule is easy to estimate for jobs without reference video.
- Limits: Input video changes both the rate and billable duration. The supplied notes list higher resolutions for Fast, while other services describe narrower Fast support; verify the exact endpoint’s accepted resolution before depending on it.
Best for: Teams that want explicit model-specific endpoints and detailed multimodal request fields.
9. Seedance API
Seedance API offers a focused REST service with a fixed set of duration and resolution choices. An agent posts to /v2/generate, stores the returned task ID, polls /v2/status, and reads the final MP4 URL after success. A callback URL is also available.
- Features: The supplied material lists text-to-video and image-to-video, a prompt of up to 2,000 characters, one or two image URLs for starting and optional ending frames, and 480p, 720p, or 1080p output. Duration choices are 5, 10, or 15 seconds.
- Pricing: Pay-as-you-go credits convert at 100 credits per dollar in the supplied pricing notes. A five-second 720p standard clip is 240 credits, or $2.40; Fast is 200 credits, or $2.00. Monthly subscriptions add bonus credits and concurrency allowances.
- Advantages: Fixed choices are simple to validate in an agent tool. The callback and status endpoint cover the normal asynchronous lifecycle.
- Limits: Fixed durations leave less room for a precisely timed 7- or 12-second shot. Subscription bonus credits can change the effective cost relative to the pay-as-you-go example.
Best for: Teams that prefer a small, predictable parameter set and optional subscription capacity.
10. APIPod
APIPod describes a unified access layer for Seedance modes, including text-to-video, image-to-video, reference-to-video, editing, and extension. Its supplied documentation emphasizes full multimodal input and synchronized sound, while using token accounting under the hood.
- Features: The documented reference caps are nine images, three videos totaling up to 15 seconds, and three audio clips totaling up to 15 seconds. It lists 4-to-15-second output, native stereo sound, first/last frames, and standard-model resolution options reaching higher tiers.
- Pricing: The notes give reference rates of $0.063 per second for standard and $0.045 for Fast, which imply about $0.315 and $0.225 for a simple five-second calculation. Separate token-based examples vary by resolution and input mode. A job with reference video can bill input and output video volume.
- Advantages: The same general API shape across model variants can simplify switching after an agent finishes drafting. The reference limits are specific enough to validate uploads.
- Limits: The supplied per-second and per-resolution examples do not always reconcile. Obtain a configuration-specific quote before using a single headline rate as a budget rule.
Best for: Teams that need several Seedance task types and are comfortable checking token-based quotes for each configuration.
Key Takeaways
- Native MCP support is the exception in this set. Monid explicitly lists an MCP route. The other documented services can still be called by an agent through REST, an SDK, or a custom MCP wrapper, but that wrapper needs to manage job IDs, polling or callbacks, errors, and result URLs.
- A five-second price is only comparable when the settings match. A 480p Fast draft, a 720p standard output, an upscaled 1080p file, and a job with five seconds of reference video do not have the same billing basis.
- Reference video can change the entire formula. Several providers charge a lower per-second rate when video is supplied but count both source and output duration. An agent should calculate the full job cost before submitting it.
- Choose controls for the actual shot. First/last-frame support helps with a planned transition. Asset tagging matters for a scene with several visual and audio references. Returned last frames can help chain clips.
- Plan for asynchronous work. A useful agent tool returns the task ID immediately and provides a safe way to resume when the video is done. Webhooks are helpful when the agent orchestration system can receive them; polling works when it cannot.
Conclusion
There is no single best Seedance 2.0 API for every agent. Monid stands out when native MCP access is the main requirement. ApiPass offers detailed input controls and a choice of service channels. Runware and APIPod cover broad video tasks; Replicate and Seedance API make individual job pricing relatively easy to understand; OceanToken adds a meaningful render-mode choice. OpenRouter, APIMart, and WaveSpeed each offer a useful API pattern for teams already working in their ecosystems.
Visit this website: Top 10 All-in-One AI Image Platforms That Bundle Multiple Models in 2026
Start with the agent’s actual task: the required assets, output length, resolution, delivery method, and monthly volume. Then compare live quotes for that same job, run a small set of representative prompts, and keep the provider whose output and task behavior fit the workflow.