～關於我的文章規劃～

neovim 的絲滑之旅

近兩個月來從 vim 轉到 neovim，彷彿有從 Windows 轉到 Linux 的豁然開朗 (X)。可能包含怎麼從 0 開始到使用 neovim 很絲滑，亦或是常見的快捷鍵整理，也可以包括我最近研究的 LaTeX + neovim 的絲滑組合，彷彿在 local 端有了 overleaf 般的舒暢，都是可以談論的範圍。

KKBOX 實習分享

這個欠超久了...之前都只有簡單寫 FB、Linkedin，我從中真的獲得很多也非常開心，但反而使我覺得不管怎樣寫都不夠好，結果就一直拖延「正式」實習心得文了QQ

字幕生成系統是怎麼煉成的

關於這個是從我跟同學在台大 DSP 課程期末專題誕生的產物，起初希望可以偷懶，把上課影片都扔進去跑一次，直接讀字幕快速上課。但之後發現使用的 WeNet Toolkit 無法應付中英混雜，加上教授授課都習慣中英混雜。近期串接 Whisper 模型，正確率高得嚇人，包括老師說「老師把毛衣脫起來一下」都正確辨識無誤，中文、英文都辨識的很好，最後也拿來輔助我們期中考準備。之後想做一個 side project，規劃字幕生成系統的前後端，同時因為 Whisper 是需要一定程度的算力的，希望可以規劃一個系統可以控制流量XD 而我相信也會是一個很好用的工具

...

A Smooth Ride with Neovim

Over the past two months, I've moved from vim to neovim. It feels a bit like switching from Windows to Linux and suddenly seeing the light (X). I could write about getting started from scratch and making neovim feel smooth to use, put together a list of handy shortcuts, or share the LaTeX + neovim setup I've been exploring lately. That combination feels as comfortable as having Overleaf running locally. Any of those could make a post.

Reflections on My KKBOX Internship

I've owed you this one for ages... So far, I've only written a little about it on FB and LinkedIn. I got so much out of the experience and had such a great time that I keep feeling no write-up could do it justice. Which, of course, is how I've managed to keep putting off the "proper" internship post. QQ

The Making of a Subtitle Generator

This started as the final project I worked on with classmates for our DSP course at NTU. The original goal was to take a shortcut: feed in all the lecture videos in one go, then read the subtitles to get through the material faster. It turned out that WeNet Toolkit couldn't handle Chinese and English mixed together, though, and our professors switch between the two all the time. I recently hooked it up to Whisper, and the accuracy was almost scary. It even got the professor's "Let me take my sweater off for a moment" right. It handles both Chinese and English really well, and we ended up using it to help prepare for our midterms. I'd like to turn this into a side project next, with a frontend and backend for generating subtitles. Whisper does need a fair bit of computing power, so I'd also like to work out how the system should manage incoming traffic XD. I think it could be a really handy tool.

...