Yo, check this out — new trick: Seedance 2.0 in ComfyUI now lets you generate videos from portraits. Works for both AI-generated virtual avatars (ones that don’t exist IRL) and real people. Just upload one image and you get a video with consistent identity and natural expressions.
Gotta break down the difference though. For virtual portraits — AI-generated, no real person behind it — no verification needed. Just drop the Seedance 2.0 node in like a regular video node, feed it your AI character image, and you’re good. ByteDance’s server will auto-detect the image type; if it flags it as protected content, it’ll throw an error, but if it passes, you’re free to use it. Real people are more of a hassle: first time you run it, you need to verify. Upload the portrait, the workflow generates a verification link. Open it on your phone or browser, do a quick <30-second liveness check, and once you pass, you get two IDs — Group ID (this identifies the person, save it) and Asset ID (this is for that specific image, can use directly). After that, new images of the same person don’t need re-verification. Just upload the new photo, plug in the saved Group ID into the node, and the system compares facial features — if it matches, it auto-activates.
This verification is ByteDance’s doing, to prevent people from misusing others’ portraits, and also to keep up with the emerging AI transparency regulations. For someone like me who works on character consistency, the best part is that across different shots, the face and performance stay solid throughout the whole clip.