Basic text-to-video, image reference, first/last frame, and video reference
workflows still use the Seedance 2.0 Video Generation
API. This page only covers the
extra asset flow expected for real-person and virtual human assets.
Capability Status
| Capability | Current status | Notes |
|---|---|---|
| Basic text-to-video / image-to-video | Available through the base Seedance API | Create tasks with /seedance/api/v3/contents/generations/tasks |
| Real-person assets | Compatibility in development, waiting for launch | Expected to require authorization, verification, and asset status checks |
| Virtual human assets | Compatibility in development, waiting for launch | Expected to use a virtual human asset library or authorized asset ID |
| Real-person + virtual human combined tasks | Compatibility in development, waiting for launch | Expected to reference multiple verified character assets in one video task |
Flow Overview
Prepare assets and authorization
Confirm that the real-person likeness, voice, styling, or virtual human
identity is legally authorized. Real-person assets are expected to require
authorization and verification; virtual human assets are expected to require
asset library usage permission.
Create or select character assets
A real-person verification flow is expected to produce a reusable asset ID
after approval. Virtual humans are expected to be selected or created from
an asset library. Field names and creation paths are not fixed before
launch.
Submit the Seedance video task
The video task will still use Seedance parameters such as prompt, model,
duration, ratio, and resolution. Character assets will be passed as
additional references. The final request schema will be defined in the
launch documentation.
Real-Person Asset Flow
Real-person assets are for workflows that need to keep a specific person’s likeness, motion style, or character consistency. After launch, the flow is expected to include:- Upload or submit real-person source material.
- Complete authorization and the required verification flow.
- Wait until the asset becomes available.
- Reference the real-person asset in a Seedance video task.
- Query the video task and download the result.
Virtual Human Asset Flow
Virtual human assets are for digital humans, brand characters, virtual hosts, and enterprise spokesperson workflows. After launch, the flow is expected to include:- Select an authorized asset from the virtual human asset library, or create one through the platform workflow.
- Obtain the character asset ID for video tasks.
- Describe the virtual human’s action, camera movement, speech, or scene in the prompt.
- Reference the virtual human asset ID in the Seedance video task.
- Query the task and download the result.
A virtual human asset does not bypass real-person authorization. Virtual
characters based on real people still need to satisfy likeness, voice, name,
trademark, and related authorization requirements.
Real-Person + Virtual Human Combined Tasks
When a video needs both real-person and virtual human assets, the expected shape is “multiple verified character asset references + one Seedance task”:| Stage | Real-person asset | Virtual human asset |
|---|---|---|
| Pre-check | Authorization, verification, asset status | Asset library permission and asset status |
| Task reference | Reference the real-person asset ID | Reference the virtual human asset ID |
| Prompt control | Describe action, expression, and camera position | Describe action, interaction, and scene |
| Result retrieval | Query and download with the same task ID | Query and download with the same task ID |
Integration Preparation
- Prepare real-person authorization, asset ownership, virtual human copyright, and allowed usage scope before integration.
- Design the client flow as separate steps: asset creation, asset selection, asset status check, and video task submission.
- Do not hard-code unreleased field names, endpoint paths, or asset ID formats before launch.
- After launch, test with short, low-resolution jobs before expanding into production workflows.
Before launch, this public document only describes the workflow. When the
capability is released, this page will add callable endpoints, request
examples, response fields, error codes, and a complete Python example.