Skip to main content
Glama
skeetmtp

byteplus-seedance-mcp

by skeetmtp
OverviewSchemaSourceRelated ServersScoreDiscussionsAdmin
Python
Local

Create a Seedance video generation task

seedance_create_video

Start a video generation task on BytePlus ModelArk and receive a task ID for asynchronous polling. Supports text-to-video, image-to-video, and reference-based video creation.

Instructions

Start a Seedance video generation task on BytePlus ModelArk and return its task ID. Generation is ASYNCHRONOUS and takes minutes: this tool returns immediately without a video. Follow it with seedance_wait_for_video (blocks until the video is ready - use this when the user just wants the video) or seedance_get_video (single status check).

Choose exactly ONE input shape; BytePlus rejects mixtures:

  1. text-to-video: prompt only.

  2. image-to-video: first_frame (+ optional last_frame) - the output starts (and ends) on those exact frames.

  3. omni reference-to-video: image_references / video_references / audio_references - the model borrows subjects, style, motion or sound from them. Seedance 2.5 accepts up to 30 images, 10 videos and 10 audio clips, and can run on audio alone. Refer to them in the prompt as '@Image 1', '@Video 2' etc. first_frame/last_frame cannot be combined with the *_references parameters.

Media may be a local file path, a public https:// URL, or an asset://. Local images (jpeg, png, webp, bmp, tiff, gif, heic, heif; max 30 MB) and local audio (wav, mp3; max 15 MB) are inlined automatically. Local VIDEO files cannot be uploaded - BytePlus takes reference video only as a public URL, so host it first.

Seedance 2.5 supports resolution 480p/720p/1080p, ratio 16:9/4:3/1:1/3:4/9:16/21:9/adaptive, and duration 4-30 seconds (or -1 to let the model choose). It does NOT support seed or camera_fixed - describe camera moves in the prompt instead.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
seedNoRandomness seed, -1 to 2147483647. Supported by Seedance 1.5 pro and 1.0 pro only - NOT by Seedance 2.5.
modelNoOverride the configured model ID or endpoint ID for this call, e.g. 'dreamina-seedance-2-5-260628'. Omit to use the server default.
ratioNoAspect ratio: '16:9', '4:3', '1:1', '3:4', '9:16', '21:9' or 'adaptive'. Default 'adaptive'. Must be 'adaptive' for first_frame/last_frame tasks and for video edit/extend tasks.
promptNoWhat to generate: subject, action, setting, camera movement, lighting, mood. Put spoken dialogue in double quotes to drive the generated audio. Keep it under ~1000 English words. Required unless media references are supplied.
durationNoOutput length in seconds. Seedance 2.5: 4-30, or -1 to let the model pick. Omit to use the model default.
watermarkNoAdd an 'AI Generated' watermark in the lower-right corner.
last_frameNoImage the video must end on. Requires first_frame.
resolutionNo'480p', '720p' or '1080p' on Seedance 2.5. Default 720p.
first_frameNoImage the video must start on (local path, https:// URL, or asset://<ID>). Mutually exclusive with the *_references parameters.
callback_urlNoHTTPS endpoint BytePlus POSTs to when the task status changes.
camera_fixedNoLock the camera. Supported by Seedance 1.5 pro and 1.0 pro only - NOT by Seedance 2.5, where camera behaviour belongs in the prompt.
service_tierNo'default' for online inference, or 'flex' for cheaper offline runs.
generate_audioNoWhether the output carries synchronised sound (voices, effects, music). Defaults to true on Seedance 2.5; pass false for a silent video.
audio_referencesNoReference audio (wav/mp3) to drive speech, music or timing. Seedance 2.5: up to 10, each 2-30s, 30s total, and audio-only input is allowed.
image_referencesNoReference images for an omni reference-to-video task (subject, character or style guidance). Seedance 2.5: up to 30. Cite them in the prompt as '@Image 1'.
video_referencesNoReference videos, as public https:// URLs or asset://<ID> only (local video files cannot be uploaded). Seedance 2.5: up to 10, each 2-30s, 30s total. Cite them in the prompt as '@Video 1'.
return_last_frameNoAlso return the final frame as a PNG, so it can seed the next clip's first_frame for a continuous sequence.
omni_reference_task_typeNoSeedance 2.5 only, and only with reference media. 'edit' to modify a reference video (requires ratio 'adaptive' and duration -1), 'extend' to continue one (requires ratio 'adaptive'), or 'auto'. Setting it moves validation to submit time instead of failing asynchronously mid-generation.

Output Schema

TableJSON Schema
NameRequiredDescriptionDefault
modelYes
statusYesTask state at submission time; always pending server-side.
requestYes
task_idYes
next_stepYes

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observedv0.1.0

TDQS

A4.6/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Without any annotations provided, the description carries full burden and does well: declares asynchronous gen, delays minutes, returns task ID immediately. It explains that BytePlus rejects mixed input shapes, details media upload rules (local images inlined, local videos not supported), and notes model-specific behaviors (Seedance 2.5 doesn't support seed/camera_fixed). Minor gap: doesn't describe error behavior if a reference is invalid.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Description is well-structured with clear sections: async behavior first, then three input shapes, then media constraints, then parameter notes. It front-loads the most critical info (async return, sibling tool guidance). It is somewhat long but every sentence earns its place for a complex tool with 18 parameters. Could be slightly tighter by combining repeated mentions of model version constraints.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a high-complexity tool with 18 parameters and no annotations, the description covers all essential aspects: the three input shapes and mutual exclusivity, the async behavior, sibling tool workflow, media upload rules per type, model-specific limitations (Seedance 2.5 param constraints, max counts), citation syntax, and special parameters (return_last_frame chaining). The output schema exists, so return values are covered separately.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100% so baseline is 3; the description adds significant value by grouping parameters into three mutually exclusive shapes, explaining relationships (first_frame mutually exclusive with *_references), detailing citation syntax (@Image 1, @Video 2), specifying upload constraints per parameter type, and clarifying model-specific constraints (seed, camera_fixed not for 2.5). Some schema descriptions already include individual parameter details, but the description's structural insight exceeds schema alone.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states it starts a Seedance video generation task, returns a task ID, and explicitly distinguishes the asynchronous nature from sibling tools like seedance_wait_for_video. It precisely describes three input shapes (text-to-video, image-to-video, omni reference-to-video) with exact parameter combinations, making it unambiguous what the tool does.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description provides explicit guidance on when to use this tool vs. alternatives: states it is asynchronous and returns immediately, then instructs agents to follow with seedance_wait_for_video (when user wants video) or seedance_get_video (single check). It also explains that media fields have different constraints (local images vs. public URLs for videos), which directly affects tool selection and parameter setup.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Install Server

Other Tools