Seedance 2.0 Real Human Face Rules 2026: Official Limits

One of the most important Seedance 2.0 questions in 2026 is also one of the most poorly explained:

Can Seedance 2.0 generate videos from real human faces?

The short answer is:

Not in the simple, default, globally open way many older guides imply.

Across current official materials from CapCut, BytePlus, and Volcengine, the consistent signal is:

• real-human-face workflows are still more restricted
• virtual-character workflows are the safer default path
• route, region, and product surface matter a lot

If your product depends on identity-locked avatars, creator likeness, or real-person face references, this is not a detail you can leave vague.

Related: Start with the new Seedance availability guide, then read the current Seedance API guide and Seedance pricing guide.

TL;DR

As of April 3, 2026, the safest public reading is:

• BytePlus docs: real human images are not supported in the standard Seedance 2.0 flow there
• CapCut official rollout: the launch included safeguards around people's likeness, and some accounts were restricted from generating from images or videos containing real faces
• Volcengine docs: direct upload of reference images or videos containing real human faces is not supported in the virtual-portrait workflow; official docs push users toward a trusted virtual portrait library

So if you need a public, documented default answer, it is:

Seedance 2.0 currently favors virtual-character routes over real-human-face routes.

What BytePlus Officially Says

BytePlus's official Digital character library docs are unusually explicit.

They say that:

• Seedance 2.0 is available there only through Model Playground within the free quota
• it does not support API invocation at that time
• generating videos using real human images is not supported
• users are recommended to use virtual characters provided by the platform

That is already enough to rule out one common bad assumption:

"If Seedance is visible on BytePlus, then real-person workflows must be generally open."

That is not what the official docs say.

The safer takeaway is:

• the public-facing global route is constrained
• the default supported route is virtual-character based
• real-human-face usage is not the baseline public workflow

What CapCut Officially Says

CapCut's official newsroom adds a second layer of confirmation.

In its April 1, 2026 Dreamina Seedance 2.0 launch note, CapCut said the rollout included safeguards against:

• unauthorized use of people's likeness
• unauthorized use of intellectual property

More importantly, CapCut said the initial rollout would restrict certain capabilities for some accounts, including:

• the ability to make videos from images or videos that contain real faces

That matters because it confirms the issue is not only theoretical or hidden in legal text.

CapCut is explicitly framing real-face input as a restricted safety-sensitive capability in the public rollout.

So even when Seedance is available inside consumer products, the real-person path still should not be treated as universally open.

What Volcengine Officially Says

Volcengine's official Virtual portrait library docs point in the same direction, but with more operational detail.

The docs say:

• the Seedance 2.0 series does not support directly uploading reference images or videos containing real human faces
• the experience center provides a trusted virtual portrait library for Seedance 2.0 video generation
• each virtual portrait has an asset ID
• in API usage, you can pass that asset via a URI

Volcengine also explicitly says:

• currently it is not supported to generate video using real-person images
• users are advised to use platform-provided virtual portraits

This is important because it shows the China-side enterprise and playground story is not "anything goes."

Instead, the official product path is:

1. use trusted virtual portrait assets
2. reference them with asset IDs / URIs
3. build within that compliant workflow

The Important Nuance Most Posts Miss

The right question is not just:

"Does Seedance support faces?"

The better question is:

"Which kind of face workflow, through which route, with which approval model?"

Today there are at least three different things people mix together:

That distinction matters more than the generic phrase "supports avatars."

What You Can Safely Promise to a Team

If you are writing documentation, buying software, or planning an implementation, these are the safe statements:

• Seedance 2.0 does support character-driven workflows
• the clearest current public path is through virtual characters / virtual portraits
• real-human-face workflows are still more restricted and should not be assumed open by default
• route and region matter, especially when comparing consumer rollout with enterprise access

These are the unsafe statements:

• "Seedance supports real-person avatars by default"
• "You can upload any human face and generate normally"
• "If it works in CapCut, the same face workflow must be open in the API"

The Safer Workflow If You Need Character Consistency

If your real goal is character consistency, not celebrity-likeness or identity-locking, the safer current workflow is:

1. use platform-provided virtual characters when possible
2. use approved virtual-portrait assets and asset IDs where available
3. build your product around character continuity, not around unrestricted real-person likeness
4. confirm any face-sensitive enterprise requirement before productizing it

This often gives you the actual creative benefit you want without relying on unsupported or compliance-heavy assumptions.

Who Should Care Most

This topic matters especially for:

• teams building avatar or spokesperson products
• agencies promising client likeness workflows
• developers building video automation on top of Seedance
• buyers comparing public consumer access with China enterprise routes

If you are only making abstract B-roll or product demos, this restriction matters less.

If you are building face-centric products, it matters a lot.

FAQ

Can Seedance 2.0 generate videos from real human faces?

You should not assume that by default. Current official materials from BytePlus, CapCut, and Volcengine all point to stronger restrictions around real-human-face workflows.

Does Seedance 2.0 support avatars?

Yes, but the clearest public route is through platform-provided virtual characters or virtual portraits, not unrestricted real-person face uploads.

Can I use the API with virtual characters?

On the Volcengine side, official docs say virtual portraits can be passed into API workflows via asset IDs and URIs. On the BytePlus global playground side, the docs still said API invocation was not supported there at that time.

If Seedance works in CapCut, does that mean face workflows are open everywhere?

No. CapCut consumer rollout and API availability are different issues, and CapCut itself described restrictions around content with real faces.

What is the safest route for face-sensitive projects?

Assume virtual-character routes are the public default, and confirm any real-person likeness workflow directly with the vendor before you design product promises around it.

Official Sources

• CapCut newsroom: Dreamina Seedance 2.0 launch
• BytePlus docs: Digital character library
• Volcengine docs: Virtual portrait library

Next Steps

• Check where Seedance is actually available: read Seedance availability guide
• Check the current API path: read Seedance API guide
• Check the pricing and route split: read Seedance pricing