Vibing Vibing The Intelligent Voice Input Method 智能语音输入法

Vibing can do three things, um, first it removes filler words like um and uh, second it handles auto-corrections — no, self-corrections, and uh, third it like auto-formats lists

Vibing can do three things:
1. Remove filler words
2. Handle self-corrections
3. Auto-format lists

Vibing 能做三件事，呃，第一个是去掉嗯啊之类的填充词，第二个是处理口误自我更—不对，自我纠正，嗯第三个就是自动把列表格式化

Vibing 能做三件事：
1. 去除填充词
2. 处理自我纠正
3. 自动格式化列表

Yesterday we discussed three issues on Teams: the UI polish, 模型那边的 latency optimization, and the prompt tuning

Yesterday we discussed three issues on Teams: UI polish, latency optimization on the model side, and prompt tuning.

昨天在 Teams 里我们讨论了三个 issue，一个是 UI 怎么做得更漂亮，一个是 latency 怎么从 model 层面去解决，还有一个是 prompt 的优化

昨天在 Teams 里我们讨论了三个 issue：一个是 UI 怎么做得更漂亮、更吸引人；一个是 latency，怎么从 model 层面去解决；还有一个是 prompt 的优化。

OK so basically what I want is, um, a voice input app, right, and it needs to like work system-wide on both Windows and Mac, and uh the key thing is it should use an LLM to clean up what you say, you know, remove filler words and fix grammar, and also it should be context-aware, like it reads your screen to figure out if you're in an email or a chat or a doc, and then format the output accordingly, oh and also it needs to support hotwords for technical terms, and um mixed language input like Chinese and English together, and I guess one more thing, it should be really fast, like the latency needs to be low enough that it feels real-time

Build a system-wide voice input app (Windows + Mac) with the following requirements:

1. LLM-powered text cleanup — remove fillers, fix grammar, keep intent
2. Context-aware formatting — read screen to detect app context (email / chat / doc) and adapt output
3. Hotword support for technical terms and proper nouns
4. Mixed-language input (e.g., Chinese + English)
5. Low-latency pipeline — output should feel real-time

我想做一个语音输入法，就是，它要能在Windows和Mac上全局使用，嗯，然后最关键的是要用大模型来整理你说的话，就是那种，去掉语气词、修正语法，还要能感知上下文，比如说，它会截屏来判断你是在写邮件还是聊天还是写文档，然后自动调整输出格式，对了还有，要支持热词来识别专业术语，嗯，还有中英文混合输入，我想想，还有一点，延迟要低，要像实时一样

开发一个全局语音输入法（Windows + Mac），具体需求如下：

1. 大模型文本清理——去除语气词、修正语法、保留核心意图
2. 上下文感知格式化——截屏识别当前应用（邮件/聊天/文档），自动调整输出格式
3. 热词支持：识别专有名词和技术术语
4. 中英文混合输入
5. 低延迟管线：输出需要达到实时体验

Yeah so basically my project needs a Mac for development, I'm building a new voice input experience, I'd like to borrow one for about three months

Hi IT Support,

My project requires a Mac for development — I'm building a new voice input experience. I'd like to borrow one for approximately three months.

Thanks,
[Your Name]

我的项目是基于语音识别来做一个全新的输入法体验，需要用 Mac 来开发，希望能借一台 Mac Studio，大概三个月

Hi IT Support,

My project is building a new voice input experience based on speech recognition, and I need a Mac for development. I'd like to borrow a Mac Studio for 3 months.

Thanks,
[Your Name]