yeko @LivLiv

個人開発者? / producer of project LivLiv @livliv_ai

also made

[yeko @LivLiv @yeko_L]

@Ainzx442917 @livliv_ai feel free to use the player, it's Apache-2.0. as for the characters, you can email [email protected] or DM me

[yeko @LivLiv @yeko_L]

We open-sourced LivLiv Player, a 2.5D web character player, ~34 KB gzipped, with lip sync and webcam face tracking. Easy to embed in your site or app.

We also open-sourced our 21 featured characters, carefully made with @livliv_ai , are ready to use.

https://t.co/wdB2VMIA6V

[yeko @LivLiv @yeko_L]

@lalunemoe 我实际的体验是 openspec 随着开发进程漂移严重，因为SDD本质上是对于模型能力的约束，而现在模型能力很强了，try Matt Pocock's Skills

[yeko @LivLiv @yeko_L]

@VictorKay97 @livliv_ai The player never asks on load: the camera only starts when the page calls startFaceTracking, usually from a button.
Without it, the character still blinks, breathes and follows the pointer, so saying no costs little.
No built-in pre-prompt UI yet, though. Good idea.

[yeko @LivLiv @yeko_L]

@embergall_ @livliv_ai https://t.co/rhQ4G9RvTp
try the demo
VRM 0.x and 1.0 load. A Blender export should work, but we haven't tested one yet, so tell us if it breaks.

[yeko @LivLiv @yeko_L]

@kachersoft @livliv_ai @bot @bot check this!!!

[yeko @LivLiv @yeko_L]

@timeblind_ @ljsabc 我们热更新播放器策略，应该对于所有角色都能生效，你再试试~

[yeko @LivLiv @yeko_L]

向您介绍 Nana (奈々), Project LivLiv 的看板娘。

在过去的几周里，我在算法和工程层面进行了许多工作，是为了把Nana，以及您想要创造的角色带到虚拟与现实的边界里。

多亏了 See-through (谢谢 @ljsabc ) 、StdGEN 等工作，我们有幸能够继续优化工程算法，来把可用的角色生成控制在一分钟以内，并建立效果评测和自动化优化方法来持续优化。

您可以直接在我们的产品中体验该角色Demo，支持在网页中进行实时面部动捕，手感参数调整等: https://t.co/ORV5KIeqxL

欢迎关注我们的官方帐号 @livliv_ai ，我们会在近期开源优化方法。

[yeko @LivLiv @yeko_L]

@timeblind_ 谢谢体验，反馈有收到，我们这几天正在加紧优化线上版本效果。

[yeko @LivLiv @yeko_L]

LivLiv にご関心をお寄せいただき、誠にありがとうございます
Thank you for your interest in LivLiv.

[yeko @LivLiv @yeko_L]

@notcillaj 感谢您提问。2D 场景的关键问题是缺少的信息太多了，我们有系统实验过让 Agent 调整绑定、用视频拟合形状键，但是效果都不是很好。参数化模板目前还是效果还是更好，成本速度都更可控。

[yeko @LivLiv @yeko_L]

@0xnuur 演示角色有经过额外的细节调整（但仍是自动化流程）, 重建算法相比线上的版本会领先一个版本
另外您也可以点击下面的Demo链接, 支持直接生成新角色（登录需要)

[yeko @LivLiv @yeko_L]

@Hitomiiim 效果真的非常棒！我是 https://t.co/Z3r3wXpGQE 的产品经理，在 livliv 里我们出于工程效率的考虑，结合 see-through 的思路使用了启发式的差分生成，如果有机会可以交流！

[yeko @LivLiv @yeko_L]

删除了所有的推文之后 似乎我的账号的推文显示完全混乱了...不过没关系，欢迎来到长毛象🐘

[Steven Cheng @stevencheng]

Just found an interesting open-source tool called sub2sub. It's made by one of my friend Meko.

It’s not really about remote desktop.

It solves a more interesting problem:

What if you have multiple computers, multiple Claude Code / Codex instances, different models, and different environments — and want them to work together?

For example:

Your Windows PC is running Codex and handling the main project.

But a specific task is better suited for Claude Code, which is already configured on your Mac.

Normally:

Copy the code → switch machines → run the other agent → move the result back.

With sub2sub, you can turn the AI agent on another machine into a remote worker:

Codex → sub2sub → Claude Code on another machine → execute → return the result

And you can continue the conversation from the original agent.

Some interesting use cases:

• One machine handles the main project, another handles refactoring
• Run Claude Code on Mac and Codex on Windows
• Route different tasks to different AI models
• Reuse different development environments across machines
• Run multiple AI coding agents in parallel
• Build a “virtual dev team” as a solo founder / OPC

What I find interesting isn't the remote execution itself.

It’s this:
An AI agent no longer has to equal one computer.

Codex can be the manager.
Claude Code can be a worker.
Another machine can host another worker.

And suddenly:

One person + multiple machines + multiple AI agents = a small AI team.

This is exactly the kind of infrastructure that could make the solo-founder / OPC model much more powerful.