Can you make money live streaming with an AI avatar? (the Neuro-sama case)

After the faceless YouTube channel, we reach the most extreme format in this entire cluster: an AI avatar live streaming with no human behind it controlling the character in real time. Here there's really only one reference case, and it's so large it distorts every expectation — it's worth separating the real case from what can actually be built today.

What an AI avatar in streaming is (and what it isn't)

There's a distinction almost nobody gets right: a traditional VTuber (Ironmouse, most of Hololive, TheBurntPeanut) is a real person talking and acting in real time behind a virtual avatar — the streaming is 100% human, only the face changes. An AI avatar like Neuro-sama is something completely different: a language model generates the responses, a text-to-speech engine converts them to audio, and no human is acting the character in real time. This article is about the latter — the much rarer, much harder to replicate case.

The Neuro-sama case: the real numbers

Neuro-sama, created by the programmer known as Vedal and streamed on the channel Vedal987, is in 2026 the most-subscribed paid channel on all of Twitch: 165,268 paid subscribers, generating at least USD 400,000 a month from subscriptions alone — before donations, sponsorships and ads. On January 4, 2026 it hit level 126 on Twitch's Hype Train, adding over 262,793 paid subscribers in the event, equivalent to an estimated USD 1.5 million in platform spending. Vedal has been running the channel as a full-time job since 2023.

Why this case isn't replicable with an off-the-shelf tool

Here's the real difference from every earlier format in this cluster: Suno, ElevenLabs, or AI video generators are subscription products you buy and use within days. Neuro-sama isn't that — it's a custom-trained and fine-tuned language model, with a voice, memory and real-time interaction pipeline built over years of actual software development. There's no "buy Neuro-sama" button; there are years of engineering.

Accessible tools for someone without that level of development

For anyone wanting to attempt a more modest version, open-source projects do exist: Open-LLM-VTuber and AITuber OnAir let you connect an LLM (via API or local models) to a Live2D avatar with voice and chat interaction. The real requirement is knowledge of Python, APIs, and system integration — it's not plug-and-play. Reported cost runs USD 50-200 a month in API calls alone, before building anything, plus USD 200-2,000+ if you commission a custom character model.

Don't confuse "VTuber" with "AI"

Worth repeating because it causes a lot of confusion: streamers like Ironmouse (2.45 million Twitch followers) or TheBurntPeanut (74.53 million hours watched in Q1 2026) are real people with a virtual avatar, not artificial intelligence generating the content. Their audience figures don't apply as a reference for what a 100% AI-driven channel like Neuro-sama would cost or earn.

So, is it worth it?

Of every business in this digital-products-with-AI cluster, this is the only one that isn't, at its core, a content business — it's a software engineering project that also happens to livestream. The reference case (Neuro-sama) is an outlier built with years of specialized development, not a replicable template using purchased tools. For someone without that technical background, the realistic version is an experimental project using open-source tools, with audience and revenue expectations far below the case that made the format go viral.