Streaming or making video as an animated virtual avatar — a Live2D or 3D character driven in real time by a human performer's face, voice, and movement.
Last verified · 2026-07-21 · by Moe Ameen
VTubing is creating content as a virtual avatar instead of on camera as yourself. A VTuber (short for "virtual YouTuber") performs live: a webcam or headset tracks the human's face and body, and that motion drives an animated character in real time, so the avatar blinks, talks, and reacts exactly as the person behind it does. The output is a persistent fictional persona — a design, a name, a personality, a lore — fronted by a real human whose identity usually stays private.
The word matters. A VTuber is not an AI-generated character reading a script. The performer is live and unscripted the same way any streamer is; only the on-screen representation is virtual. That is the line that separates VTubing from [AI avatar video](/glossary/avatar-video), where a model synthesizes the face, the voice, and the delivery from text. VTubing swaps the *camera*, not the *human*.
Two avatar technologies dominate. Live2D rigs a flat illustration into moving layers, giving the expressive, anime-style 2D look most associated with the medium; it is cheaper to commission and lighter to run. 3D models (often built in tools like VRoid Studio and animated through Unreal Engine or game engines) allow full-body movement, dancing, and 3D spaces, at higher cost and setup complexity. Face tracking runs off an ordinary webcam or a phone's depth camera; body and hand movement need VR trackers or a mocap rig.
VTubing traces to a single channel. Kizuna AI launched "A.I.Channel" in late 2016 — the channel was created October 18, 2016, her first video went up November 29, and her self-introduction video (December 1, 2016) is where she called herself a "Virtual YouTuber," coining the term. She was a 3D character driven by real-time motion capture, and by 2018 she was mainstream enough in Japan to appear in tourism and brand campaigns.
Agencies formalized the medium fast. Cover Corporation's Hololive Production launched in 2017 and Ichikara's Nijisanji in 2018, both building rosters of "talents" — performers assigned a designed avatar and persona — and both leaning on the cheaper, faster-to-produce Live2D style that let them debut many characters. By 2020 there were already thousands of active VTubers.
The medium went worldwide in 2020. Hololive's first English branch, Hololive English -Myth-, debuted in September 2020, and one of its members, Gawr Gura, became the most-subscribed VTuber on YouTube — the first to pass four million subscribers, reaching roughly 4.7 million before graduating (retiring the character) in 2025. English-language agencies and a large independent scene followed, and "VTuber" stopped being a Japan-specific term.
| Platform | Behavior |
|---|---|
| YouTube | The historical home of VTubing and where the "virtual YouTuber" name comes from. Long live streams, superchats, and premieres. Most agency talents and the biggest subscriber counts live here; VOD clips and stream highlights drive the discovery loop. |
| Twitch | The dominant home for Western and indie VTubers. Live-first, subscription- and bit-driven, with a heavy clip culture. Avatar overlays sit on top of gameplay or Just Chatting the same as a facecam would. |
| TikTok / Shorts / Reels | Where VTubers grow. Vertical clips of stream moments — a funny reaction, a highlight, a song cover — are the top-of-funnel that sends new viewers to the live channel. Most VTuber discovery in 2026 happens in short-form, not in the live streams themselves. |
| X and image platforms | Where the persona lives between streams. Avatar art, schedule announcements, and "fan art" reposts keep the character present when the performer is offline; the visual identity is a brand asset, not just a stream overlay. |
The instructive thing about VTubing for anyone building with AI avatars is that it proves the audience was never asking for a real face — they were asking for a consistent, characterful identity they could form a relationship with. A well-designed animated persona out-earns a lot of on-camera creators precisely because the character is a durable, ownable brand asset that a bad-hair day or a face-reveal scandal can't dent.
But do not confuse the two things. VTubing's whole appeal is that a live, improvising human is behind the avatar; the parasocial bond is real because the performer is real. AI avatar video — the [Persona Shorts](/glossary/persona-shorts) and avatar tooling most content engines ship — solves a different problem: producing talking-head volume without a shoot. The overlap is only the visual shell.
Where a production engine like [Kompozy](/) is actually useful to a VTuber is off-stream, in the part almost nobody wants to do. The stream is the human's job and stays the human's job. Turning three hours of it into the twenty vertical clips, the recap post, and the schedule graphics that feed nine platforms is mechanical distribution work — and that is the layer Kompozy owns, keeping the character's voice and look consistent across every clip and caption via a [Persona Brief](/glossary/persona-brief) while the performer goes back to streaming.
VTubing is creating content — usually live streaming — as an animated virtual avatar instead of appearing on camera as yourself. A webcam or headset tracks the human performer's face, voice, and movement in real time, and that drives a Live2D or 3D character, so the avatar reacts exactly as the person behind it does. The performer is live and unscripted; only the on-screen representation is virtual.
No. Despite the name (the pioneer character Kizuna AI called herself an "AI"), a VTuber is a real human performing live through an animated avatar. That is the key difference from AI avatar video, where a model synthesizes the face, voice, and delivery from a script. VTubing swaps the camera, not the human.
It started in Japan with Kizuna AI, who launched her "A.I.Channel" in late 2016 and coined the term "Virtual YouTuber" in her December 2016 self-introduction video. Agencies followed — Hololive in 2017 and Nijisanji in 2018 — and the medium went worldwide in 2020 when Hololive English debuted and its member Gawr Gura became the most-subscribed VTuber on YouTube.
Live2D rigs a flat illustration into moving layers for an expressive anime-style 2D look; it is cheaper to commission and lighter to run, which is why most VTubers use it. 3D models, often built in tools like VRoid Studio and animated in a game engine, allow full-body movement, dancing, and 3D spaces at higher cost and more complex setup requiring VR trackers or a mocap rig.
At minimum: an avatar model (a Live2D rig or a 3D model), face-tracking software that runs off an ordinary webcam or a phone's depth camera, streaming software, and a decent microphone. Full-body 3D VTubing additionally needs VR trackers or a motion-capture setup. Many people start with a free or template model and upgrade the rig as the channel grows.
Usually not. The persona is designed to keep the performer's identity private, which is part of the appeal — the character is an ownable brand that is insulated from the individual. Anonymity is not absolute, though: voice, schedule, and metadata leaks have de-anonymized performers, so operational security matters.