Lesson 01
The three kinds of “AI streamer”
The AI VTuber: a character driven by a language model, speaking through text-to-speech, shown as an animated avatar, reacting to live chat. Interactive, real-time, cheap to run; Neuro-sama is the reference and the only one of the three an individual can build today that audiences have shown up for. The AI clone of a real person: a video model generating a photoreal version of an existing streamer — N3on tried it on Kick in September 2026, lost roughly 50,000 followers in the first hour and walked it back within a day; cloning a real person's likeness without consent also violates Kick's guidelines. The “infinite” generated-video stream: scenes generated endlessly from chat prompts — a tech demo, expensive per hour, with no character anyone bonds with. We build the first one.
Lesson 02
Why the AI VTuber model works
Viewers don't pay for technology; they pay for a character they like, on a schedule, reacting to them. An AI can be there 24/7 and greet every chatter by name. Neuro-sama's history is the syllabus: a game bot, then an LLM and a voice, a free Live2D model, a two-week ban for a hateful line, content filters, a commissioned model, then segments and collabs.
Lesson 03
The rules you're operating under
Kick's community guidelines (updated March 19, 2026) allow AI content on two conditions: AI content that mimics reality must be clearly labeled in the title or an on-screen overlay, and synthetic media may not simulate a real person's presence or endorsement without permission. Twitch and YouTube have similar rules; the same two habits keep you inside all of them. If you ever sell anything related to this, FTC rules on earnings claims and endorsements apply.
Checkpoint
You can explain, in one sentence each, why the clone and infinite-video approaches aren't your product, and you've read the AI section of Kick's community guidelines.
This module is free. Modules 1–8 are in Foundations.
See Foundations