A week ago, Microsoft released the open-source VibeVoice Model, and today Google updated Gemini Audio. In the meantime, I created MeetLingo: an instant speech to speech translation tool specifically designed for PC online meetings. The reason is actually very simple. When VibeVoice said it could delay by 300ms, I realized that the delay of voice TTS was already low enough to be really used in time sensitive scenarios such as "instant meetings", so I spent a day using Vibe Coding to create the MVP. The entire system adopts a streaming architecture: speech recognition, token translation, and TTS input are carried out while speaking, and finally, idiomatic sounds are directly output instead of waiting for a sentence to be finished before processing. Because of this, currently under the premise of a local+open source model, TTFA falls around 1000-1500ms. In the open source world that can be self shelved, modified, and embedded, it is generally above 2000ms, which is already a very competitive number (comparable to the translation speed in Google Video) To be honest, when I saw the update of Gemini Audio released by Google today, I did feel for a moment that "this idea and narrative were directly suppressed". With the big company's intervention, it was already easy to overshadow the creativity, timing, and even the presence of the new venture. So in this era, what really matters is often not hiding ideas and slowly polishing them, but whether we can quickly make MVP and put it into the real world for verification. But on second thought, since it has already been done, let's continue. MeetLingo was not limited to meetings from the beginning. Essentially, it is a low latency speech to speech pipeline that will support more languages in the future and can be used in any real-time voice scenario. The only difference is that I chose to make it an open source, local first feature, rather than a platform locked feature. Since it has already started running, let's gradually optimize it to be faster, more stable, and more user-friendly. Github and website are left in the comments section. Welcome to take a look, and feel free to click on them ⭐