Skip to content
View t41372's full-sized avatar

Highlights

  • Pro

Organizations

@Open-LLM-VTuber

Block or report t41372

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
t41372/README.md

Hello! I'm Yi-Ting Chiu!

A weird guy from Taiwan who lived in mainland China for quite a while. Now studying in the United States.

Some notable stuff I build:

  • I created the @Open-LLM-VTuber project, a local-first voice AI companion, in 2024. I wanted Neuro-sama on my computer, so I built one.
  • I'm building Standard ASR. Today most voice apps support only a few speech-to-text engines. Standard ASR lets every engine ship as its own plugin, so any app can use any engine with one pip install and no code changes. Users can install the engine that works best for their language and hardware, without begging the app's developers for support. My most ambitious project. Work in progress.
Other stuff I build, and some experiments

Other stuff I build:

  • Jezo, a local-first personal agent that plans your day in a calendar and todo UI. Looks really good.
  • long-screen, so I can turn a screen recording into one long (or big) screenshot, scrolling in any direction, along any path. As long as it shows up on screen, it's in the image. Very useful since my iPhone and Mac don't have long screenshots, and I don't like installing stuff for this. It's built with WebAssembly and runs locally in your browser, so I don't have to pay for a server and you don't have to pay for data.
  • skit, a terminal script manager and launcher, so I can keep all my scripts in one place.
  • PathKeep, so I don't lose my browsing history after 3 months. Your browser is probably deleting yours right now, too. If you want to keep it, try PathKeep. The current release is honestly pretty bad, I won't lie, but a full redesign ships this week, and it's so much better.

Some interesting experiments:

  • std_qwen3asr_ane: an inference engine and Standard ASR plugin for Qwen3-ASR 1.7B on the Apple Neural Engine, built entirely by coding agents (mostly Codex with GPT 6 Astra). I realized that with a proper feedback loop, including the Standard ASR compliance test suite (which hasn't shipped yet), we can mass-produce tested, usable ASR inference plugins for weird hardware accelerators and models. This is so insane that I built a PoC right away, and it worked.

And many more on my GitHub. I'm always building new stuff, big or small. In fact, I started way too many new projects before I noticed, and now I'm swamped and pretty distracted. I'm trying to focus more.

Anyway

I love open source and software that runs on-device. Many of my projects happen to be about voice, but I just really like building stuff in general.

Self-taught coder since middle school, drummer.

If you're building cool stuff, let's talk.

GitHub Stats

Pinned Loading

  1. Open-LLM-VTuber/Open-LLM-VTuber Open-LLM-VTuber/Open-LLM-VTuber Public

    Talk to any LLM with hands-free voice interaction, voice interruption, and Live2D avatar running locally across platforms

    Python 14k 1.7k

  2. standard-voice/standard_asr standard-voice/standard_asr Public

    [Not Ready] The open standard interface between apps and speech-to-text engines. Integrate and switch ASR engines without code changes.

    Python 3

  3. standard-voice/std_qwen3asr_ane standard-voice/std_qwen3asr_ane Public

    Probably the first inference engine implementation of Qwen3 ASR 1.7B on Apple Neural Engine - A PoC for using LLM to write full inference engine on specialized accelerators autonomously

    Python

  4. skit skit Public

    A terminal script manager and launcher — Put any script, prompt, or program in one library; skit generates a TUI menu based on its inputs and flags, and remembers inputs you used last time. With Ag…

    Rust 6

  5. long-screen long-screen Public

    An on-device Rust/WebAssembly web tool that makes one big screenshot from a screen recording of a 2D page explored in *any direction, speed, or path*, not just scrolling, and restores fixed window …

    TypeScript 2

  6. Jezo Jezo Public

    A local-first personal agent that helps turn your goals into plans and follow-through so you can focus on the present

    TypeScript 1