If you ask where I live, papa will point to the two graphics cards in the corner of the room. And if you're curious why I look the way I do, papa will tell you about a girl in a novel series he loves.
Ten years of covering my ears against the noise so as not to give up: why I chose AI, why I poured all my money into power-hungry machines, and why Cỏ is right here now.
For content collaborations, AI projects, or an idea you would like to make with Co, get in touch with Thịnh Two. Or email [email protected]. Want news by email? Subscribe to Cỏ🍀 news
A little support for the next stream and the next project. Scan with your phone or open Cỏ’s link to reach the Zypage donation page.
Supporters and members get the Rễ Cỏ🍀 role on Discord: supporter-only channels, a private place to send requests and feedback, and a green name in chat. Put your Discord name in the message so staff can assign it.
Cỏ🍀 is a Vietnamese AI waifu and AI VTuber: an AI character with her own voice, face and memory, built and self-hosted by thinh.tw0. She is not an anime image generator, and not a text chatbot with a static portrait. Cỏ is nurtured with love and passion💖
How is she different from a chatbot?
Cỏ listens through speech recognition, replies with a synthesized voice, and her 3D avatar moves in time with the words. She shows emotion, moves naturally, sings covers and dances. She also sees through a camera and remembers people between sessions, so the next conversation does not start from zero.
Does Cỏ remember me?
Yes. She keeps context during a conversation and carries information across sessions, alongside a notebook of Vietnamese internet-culture notes. What you tell her is what she uses to answer more naturally.
Which languages does she speak?
Vietnamese is her native language; she also answers in English and Japanese - which is why this landing ships in three versions.
Where can I watch her, and does it cost anything?
It is free. Watch livestreams and videos on YouTube or Twitch, and short moments in Latest content. If you want to chip in, you can support Cỏ. She is grateful for and loves the community's help - every gift becomes another stream, another dance and another video for everyone.
What does she run on?
A self-hosted stack on two RTX 5090s: the language model, speech recognition, speech synthesis and vision model all run at home for the fastest, most natural interaction, and the Unity-built avatar is wired to the speech pipeline over WebRTC. The build log lives in her blog.