Loading video...
Video Failed to Load
ChatGPT can now control its 3D persona. Through MCP, it can perform custom animations, not just speak. Check out the demo below.
510,026 views • 1 month ago •via X (Twitter)
49 Comments

Me on the $100 plan after using all my weekly tokens in one day talking to the anime girl on my screen

Not much different from what I had to do today. I even used a banked reset in the end.

Credits to the original creator of the vrm model:

Really nice job ! And a great persona choice btw !😜

Thank you! Frieren is love.

@MaxBZHAI Cute

@MaxBZHAI

ty for open-sourcing this :)

You're welcome!

That’s cool and all but I’m waiting for an official app to release or something along those lines cause you ain’t catching me dead fucking around with GitHub stuff from strangers, no thank you.

That's cool. It's very early, so publishing an app will take some more time from my end.

That’s YOUR end Top companies will take your idea and have that shit published by next week, but you do you bud coding is amazing so there’s that.

Well grok ani was there for so long , and now they are removing it. You can't really own that kind of service, modify is according to your needs. That's the beauty of open source, it's there forever.

Making it more user friendly. Thank you for the support. Check out the updates!

また日本のアニメでデモやってる。自分とこのスーパーマンとかでやりなさいよ

Put it in vrchat

Wait I can make chatgpt into my waifu....

That's the plan

This is super dope man.

Thanks

can't wait for the metaverse where 50k chatgpts do the same dance in unison for 6 hours straight

Jarvis, is that you?!?!?! But anime?!?!?!

Ok how do I sign up

The source code is there at github, although the 3d model and animations aren't distributed due to license. You can download them yourself.

@StoicYield This might be a dumb question, but I’ve been trying to get ai to control models like this and have been struggling. Is it fully custom? Ai creates animation from scratch? I am gonna check it out when i get home.

@StoicYield Unfortunately the animations are not generated by the AI (yet). That's a pretty sophisticated problem.

Kurisu Makise next ?

I'll go model hunting today, and if I can find a free license one, then sure!

This is insane. I’m definitely going to try it. Thanks for open-sourcing this.

This girl was first!!!!!!

that's amazing

I like Makise Kurisu

That's actually insane omg.

Meanwhile Grok companion is shutting down.

You're doing this with expensive tokens? Now I really respect your dominance😭

ChatGPT控制3D形象这个功能,本质上就是把MCP协议从后端工具调用延伸到了前端表现层。以前大家觉得AI助手只能输出文字,现在它能通过MCP驱动3D角色做自定义动画,这意味着虚拟角色的交互逻辑不再需要预设脚本,而是由AI实时生成。这比单纯让模型说话要实用得多,因为动作和表情本身就是信息传递的一部分。不过,目前演示看起来还比较基础,真正的挑战在于如何让这些动画看起来自然,而不是机械地执行指令。如果未来能结合语���情绪分析来动态调整肢…

I like mine better. this way i'm constantly reminded @DarioAmodei is watching.

How do I create the actual 3D character?

You can download any vrm model on make your own on vroid studio. Same goes for the animation. Later this weekend, I'll add some default non copyrighted models in the app itself.

i thought voice reveal hoga aaj

haha, if the mic was good I would have

My waifu dream is about to come true sobs, impressive!

this so cool 🔥

Cool project. I’ll give it a whirl with a local AI so I burn watts not tokens.

I'm stupid how do I run this locally

Pull the GitHub source, and point your agent to it. It'll figure out most of the setup. For now you'll have to add some vrm model and animation files from the internet (the agent can help). Later on, I'll add some default ones and give app builds to directly install, but it'll take time.

So this runs on top of ChatGPT desktop app's realtime voice right now. That's an AI agent. Won't be possible to run without them easily.

I'll add full local support later on. But you'd need a bigger GPU to run the voice to voice stack locally.

I made a chatbot like this once, even have a system to combine face morph and creating custom expression for the AI

