Launch HN: Hyprnote (YC S25) – Hyprnote：一款开源的 AI 会议记录工具
19 分•作者: yujonglee•10 个月前
Hi HN！我们是来自 Hyprnote 的 Yujong、John、Duck 和 Sung (<a href="https://hyprnote.com" rel="nofollow">https://hyprnote.com</a>)。我们正在开发一款开源、注重隐私的 AI 笔记应用，完全在本地运行。可以把它想象成开源版的 Granola。没有 Zoom 机器人，没有云端 API，你的数据永远不会离开你的设备。<p>源代码：<a href="https://github.com/fastrepl/hyprnote">https://github.com/fastrepl/hyprnote</a>
演示视频：<a href="https://hyprnote.com/demo" rel="nofollow">https://hyprnote.com/demo</a><p>我们开发 Hyprnote 是因为我们的一些朋友告诉我们，他们的公司因为数据安全问题禁止使用某些会议记录工具，或者他们只是觉得把数据发送到未知的服务器上很不舒服。所以他们又回到了手动记录笔记的方式——在会议中难以集中注意力，会后又浪费时间。<p>我们问：我们能不能开发出同样有用，但完全本地化的工具呢？<p>Hyprnote 是一款桌面应用，可以在本地转录和总结会议内容。它会捕捉你的麦克风输入和系统音频，所以你不需要邀请机器人。它会根据你做的笔记生成摘要。默认情况下，所有操作都在本地 AI 模型上运行，使用 Whisper 和 HyprLLM。HyprLLM 是我们基于 Qwen3 1.7B 微调的概念验证模型。我们了解到，总结会议是一项非常微妙的任务，模型的原始智能（或权重）并没有那么重要。我们将在完成该模型的第二版（仍然不够好，我们可以做得更好）后发布更多关于评估和训练的细节。<p>Whisper 推理：<a href="https://github.com/fastrepl/hyprnote/blob/main/crates/whisper-local/src/model.rs">https://github.com/fastrepl/hyprnote/blob/main/crates/whisper-local/src/model.rs</a><p>AEC 推理：<a href="https://github.com/fastrepl/hyprnote/blob/main/crates/aec/src/lib.rs">https://github.com/fastrepl/hyprnote/blob/main/crates/aec/src/lib.rs</a><p>LLM 推理：<a href="https://github.com/fastrepl/hyprnote/blob/main/crates/llama/src/lib.rs">https://github.com/fastrepl/hyprnote/blob/main/crates/llama/src/lib.rs</a><p>我们还了解到，对于一些人来说，完全的数据可控性与隐私一样重要。因此，我们支持自定义端点，允许用户引入他们公司的内部 LLM。对于需要集成、协作或管理控制的团队，我们正在开发一个可选的服务器组件，可以进行自托管。最后，我们正在探索如何让 Hyprnote 像 VSCode 一样工作，这样你就可以安装扩展程序，并围绕你的会议构建自己的工作流程。<p>我们相信，由本地模型驱动的、注重隐私的工具将开启下一波现实世界 AI 应用的浪潮。<p>我们在这里，期待您的评论！
查看原文
Hi HN! We're Yujong, John, Duck, and Sung from Hyprnote (<a href="https://hyprnote.com" rel="nofollow">https://hyprnote.com</a>). We're building an open-source, privacy-first AI note-taking app that runs fully on-device. Think of it as an open-source Granola. No Zoom bots, no cloud APIs, no data ever leaves your machine.<p>Source code: <a href="https://github.com/fastrepl/hyprnote">https://github.com/fastrepl/hyprnote</a>
Demo video: <a href="https://hyprnote.com/demo" rel="nofollow">https://hyprnote.com/demo</a><p>We built Hyprnote because some of our friends told us that their companies banned certain meeting notetakers due to data concerns, or they simply felt uncomfortable sending data to unknown servers. So they went back to manual note-taking - losing focus during meetings and wasting time afterward.<p>We asked: could we build something just as useful, but completely local?<p>Hyprnote is a desktop app that transcribes and summarizes meetings on-device. It captures both your mic input and system audio, so you don't need to invite bots. It generates a summary based on the notes you take. Everything runs on local AI models by default, using Whisper and HyprLLM. HyprLLM is our proof-of-concept model fine-tuned from Qwen3 1.7B. We learned that summarizing meetings is a very nuanced task and that a model's raw intelligence (or weight) doesn't matter THAT much. We'll release more details on evaluation and training once we finish the 2nd iteration of the model (still not that good we can make it a lot better).<p>Whisper inference: <a href="https://github.com/fastrepl/hyprnote/blob/main/crates/whisper-local/src/model.rs">https://github.com/fastrepl/hyprnote/blob/main/crates/whispe...</a><p>AEC inference: <a href="https://github.com/fastrepl/hyprnote/blob/main/crates/aec/src/lib.rs">https://github.com/fastrepl/hyprnote/blob/main/crates/aec/sr...</a><p>LLM inference: <a href="https://github.com/fastrepl/hyprnote/blob/main/crates/llama/src/lib.rs">https://github.com/fastrepl/hyprnote/blob/main/crates/llama/...</a><p>We also learned that for some folks, having full data controllability was as important as privacy. So we support custom endpoints, allowing users to bring in their company's internal LLM. For teams that need integrations, collaboration, or admin controls, we're working on an optional server component that can be self-hosted. Lastly, we're exploring ways to make Hyprnote work like VSCode, so you can install extensions and build your own workflows around your meetings.<p>We believe privacy-first tools, powered by local models, are going to unlock the next wave of real-world AI apps.<p>We're here and looking forward to your comments!