Delulu’s ad looks like science fiction until you visit the product: choose or create an avatar, select Delulu as a virtual camera, and appear through that identity in live calls, streams, recordings, and social video. The company describes the system as a real-time identity layer rather than a filter, with sub-second processing designed to follow the user frame by frame. For anyone who has watched AI avatars evolve from static portraits into talking characters, this is the obvious next step: the avatar becomes the camera.
Where It Works
Delulu says its virtual camera can be used with Zoom, Google Meet, Teams, Discord, OBS, Streamlabs, Riverside, Loom, browser-based video chat, and other software that accepts a webcam. That makes it more flexible than a platform-specific avatar feature. Streamers can use the same identity across several destinations instead of rebuilding the character inside each service.
The Free Tier Is Short by Design
The company currently offers free transformations capped at short sessions, while paid tiers remove the per-session cap and include longer monthly usage allowances. That is a sensible way to demonstrate a compute-heavy real-time product. Anyone planning several hours of streaming should compare the included hours and hardware requirements rather than assuming the free experience scales indefinitely.
This Could Be Useful for Virtual Characters
A deliberately fictional persona can use real-time identity transformation for live Q&A, demonstrations, tutorials, or entertainment without requiring a custom 3D rig. A character like Ava Morgan is an obvious example: the visual identity already exists, so the challenge becomes keeping her appearance consistent while a human performer supplies movement and timing. That can create a much more natural live presence than pre-rendered talking-head clips.
Impersonation Is the Other Side of the Technology
The same ability to upload almost any image raises immediate consent issues. Turning yourself into an original character is different from appearing live as a real person who never agreed to it. Platforms, employers, audiences, and jurisdictions may have rules around synthetic identity, disclosure, fraud, and likeness rights. The technical fact that a camera can produce the face does not answer whether it should.
Real-Time Makes Trust More Important
Recorded deepfakes at least have a rendering pipeline and can sometimes be reviewed before publication. A live identity system removes that delay. Businesses using the technology should decide when disclosure is required, what characters are permitted, and how recordings are labeled. The more convincing the avatar becomes, the less responsible it is to rely on viewers noticing something looks synthetic.
Bottom Line
Delulu demonstrates how quickly AI avatars are moving from generated media into live presence. For streamers, fictional characters, privacy-conscious creators, and entertainment, that can be genuinely exciting. It also means impersonation no longer requires editing after the fact. The best use cases are the ones where the audience understands the character is virtual and enjoys the performance precisely because of it.
