For newbies

ComfyUI Cracks Open Real Human Faces in Seedance 2.0 — But Is It Really Fixed?

|Author: Viacheslav Vasipenok|4 min read| 9
ComfyUI Cracks Open Real Human Faces in Seedance 2.0 — But Is It Really Fixed?

Anyone who has spent time wrestling with Seedance 2.0 in ComfyUI knows the pain. The model is incredibly powerful — cinematic camera control, native audio sync, buttery motion — but the built-in censorship made generating videos with actual human faces a nightmare. You either resorted to heavy prompt gymnastics, face-swapping in post, or just gave up. Now Comfy has dropped a proper solution: two new official workflows that let you generate videos featuring real, verified people.

ComfyUI Cracks Open Real Human Faces in Seedance 2.0 — But Is It Really Fixed?The announcement came on April 23, 2026, and it’s straightforward. Comfy added dedicated Seedance 2.0 Real Human templates to the Template Library.

There are two main workflows:

  1. Reference-2-Video (R2V) — Feed it a verified portrait (or multiple references) and generate a full video driven by your prompt. 
  2. Keyframe-based generation — Control the video using two (or more) verified key frames for precise timing and performance.

The secret sauce is ByteDance’s liveness verification system, baked directly into the ByteDance Create Image/Video Asset node.

Here’s exactly how it works:

1. Drop your portrait image or short video into the node.
2. Queue the workflow — Comfy generates a one-time verification link.
3. Open the link on your phone or desktop and complete the liveness check (under 30 seconds).
4. Success. You instantly receive two IDs:
- Group ID — tied to the actual person (this is the master key).
- Asset ID — tied to this specific image or video clip.

From then on, any new photo or video of the same person only needs the Group ID. The system compares facial features and auto-approves the new Asset ID. You can now use up to 9 images + 3 videos + 3 audio clips as references in a single generation, giving you massive control over look, motion, and even lip-sync.

ComfyUI Cracks Open Real Human Faces in Seedance 2.0 — But Is It Really Fixed?Important restrictions (straight from the docs):

  • Only one person per asset.
  • Group ID is permanently bound to your Comfy account — no sharing, no transferring.
  • The verification is handled by ByteDance to comply with emerging AI transparency rules and prevent deepfake-style impersonation.

On paper, this is a huge leap forward. Identity consistency is rock-solid, expressions feel natural, and you keep full cinematic control. And yes — the liveness check does not affect the price of generations.

But the community is already sharpening the pitchforks, and for good reason.

Two big questions are circulating:

1. Face ID squatting — first come, first served?
What happens if someone beats the real owner to verification? Could a malicious (or just very fast) user upload Brad Pitt’s photo, trick the liveness check with DeepfaceLab or Wan2.1 Animate, and claim the Group ID forever? Would the actual Brad Pitt (or any lesser-known creator) then be locked out of using their own likeness in ComfyUI? The system currently has no public “claim your face” mechanism. It really does feel like “whoever verifies first owns the ID.”

2. Where the hell do you store these IDs?
Comfy is local-first. Your Group ID and Asset IDs live in the workflow JSON or in a node somewhere on your machine. If you accidentally delete the folder, reinstall Comfy, or switch computers… those IDs are gone. There’s no cloud backup mentioned, no central registry, nothing. ByteDance holds the biometric data for verification, but the actual IDs are your responsibility. Lose them and you’re back to square one.

Still, even with these rough edges, this is undeniably better than the previous “pray the model doesn’t censor your face” experience. The community will probably hack together ID managers, backup tools, and maybe even a public registry within weeks. Some will celebrate the breakthrough; others will complain it’s still too gated and bureaucratic.

Bonus update: Comfy also quietly enabled proper 1080p generation in Seedance 2.0. A full video now costs 721.5 credits — roughly $3.42 at current rates. Not cheap, but the quality jump is worth it for serious work.

ComfyUI Cracks Open Real Human Faces in Seedance 2.0 — But Is It Really Fixed?Also read:

Bottom line: Comfy and ByteDance just made real-human video generation in Seedance 2.0 actually usable.

It’s not perfect, the ID system raises legitimate concerns, and the community will have opinions (loud ones). But for the first time, you can reliably put real faces into cinematic AI video without praying the censorship gods are in a good mood.

The workflows are already live in the Template Library. Update ComfyUI (or jump into Comfy Cloud), search for “Seedance 2.0 Real Human,” and see for yourself.

Whether you greet this with excitement or side-eye… the era of verified real-human Seedance videos has officially begun.

Share:
0