hakumei

I got GPT-Live running on Yorishiro, and the voice conversation experience was interesting, so I wanted to share it.

https://github.com/sktkkoo/Yorishiro/discussions/68

This is still an experiment and is not released yet. I used the Codex app server.

Yes, you can change the model. Yorishiro currently supports only VRM 0.x/1.0. VRM is a format widely used in avatar communities, including VTubers.

Please let me know if you’d like support for other 3D model or 2D format (ex. Live2D).

Thanks!

Thanks a lot for your feedback! I think moving cost to another terminal is problem. I build Yorishiro for exploring what is possible, but I’ll consider ways to use with other terminals or editors, for example, add-on, sdk. Thanks!

I'd love to hear what you think. thanks!

Thanks! I'd really appreciate to hear your thoughts after trying it, especially from developer's perspective. I'd love to see your project too!

The draw cost is not as heavy as it looks, because it is limited to 30 fps. And I avoid unnecessary rendering when you hide the sidebar window. I’m considering performance mode which is controlled resolution and post processing effects. Thanks!

Thank you! I hope so.

Show HN: Yorishiro - macOS terminal where AI lives

Yorishiro is an open source project that Claude Code / Codex has a body like anime character. The name “Yorishiro” in Japanese means an object inhabited by spirit.

My first idea started comunicating with AI agent long time by terminal is very tired. Because AI agent is no face, no expression, no body, and I don’t see they think. So I provided a 3D body and inhabited environment to AI agent. I call it “Presence Harness”.

I devise many idea, for example, reflex function. "Aura" is white light moving instantly. It shows things to focus on before replying by LLM. When AI agent wait for your approve, the room lights flash to notify. “Yori” who is Yorishiro character express sad if AI agent encounter errors. If git push is seccess, Yori is happy face and fireworks are being launched.

Then, Yorishiro has a self reference MCP. AI agents in Yorishiro control light, body, scene, UI, effects, and so on. Yorishiro has extention supporting hot reload. Yorishiro builds itself. I love Emacs, so I make it to a self modifiable terminal.

If you want to try it,

> brew install --cask sktkkoo/yorishiro/yorishiro

I’m Japanese and not good at English. Thank you for your reading.

You're right. Italic is rarely used in Japanese, and it seems the English styling was inadvertently applied to the Japanese text as well. I'll fix it right away.

Everything is saved automatically to your browser's local storage. To export as .txt, open the Read Back screen from the top right and tap the Share button. Note: the Read Back button only appears after you've written something.

That's an interesting constraint. I also like that the blurred words stay visible — I can see how much I've written in real time. Thanks for sharing.

To clarify, I write the posts myself, carefully thinking through and refining the content. Since I'm Japanese and not a native English speaker, I use AI to assist with translation.

Thanks for the detailed feedback.

You're right about the fading — if you keep writing in one long block, lines stay. I write in short paragraphs myself, so I hadn't noticed. I'll think about how to handle this better.

On losing text: lines don't fade the moment you hit return. They wait until you start typing again. I found that fading on return alone left the screen empty at exactly the moment I needed to think.

The lag on Firefox/Ubuntu is something I'd like to look into — I haven't tested on that environment. The autocorrect issue is likely a browser limitation, but I'll see if there's a workaround.

Thank you for trying it, and for staying long enough to find the reader mode.

The frustration at the start is real. It doesn't fully go away. But for me, it slowly became something else — a feeling of gently letting go of what I'd written.

I thought hard about the reader mode. For long-form writing, an app where you can never look back is just too limiting. In the end, I decided to keep both.

Interesting concept – reminds me of Slither.io in the sense that you capture others to grow stronger. Using chess pieces as the progression system is a clever twist.

Curious whether there are plans to add forms beyond Queen – it seems like there’s room to expand the progression further.

This looks useful. The Option+C → contextual actions flow feels natural. Does it work with images on the clipboard too?