📡 Idea Radar 每天 8:00 / 18:00 推送 · 18:00 生成精选
reddit · SideProject · 56 分 ← 返回列表

I vibe-coded a Windows dictation + meeting-notes app that keeps everything on your PC. Here’s what I learned.

I’m not a professional developer. Six months ago I wanted a dictation tool that didn’t ship my audio and meeting notes to someone’s cloud, and I couldn’t find one I trusted. So I built it — Claude wrote nearly all of the code, I did the product calls, testing, and a lot of “no, not like that.” A9 Voice (Windows, free): hold a key and talk, and it types into whatever app you’re in. Sit in a meeting and it records, figures out who said what, and writes the notes. Everything lives in plain

五维评分

付费意愿
50
需求真实度
62
个人可落地
68
竞争空间
32
新鲜度
58

AI 分析

痛点:想要跨应用听写和会议纪要的用户不信任云端服务,不愿把音频和会议内容上传到别人的服务器,但市面上缺少可信的本地优先方案。

目标用户:注重隐私的知识工作者、开发者、咨询/法律/医疗等敏感行业从业者,尤其是 Windows 用户(Mac 端已有较多选择)。

证据:“Six months ago I wanted a dictation tool that didn’t ship my audio and meeting notes to someone’s cloud, and I couldn’t find one I trusted. So I built it.” “Everything lives in plain files on your PC — no account, no login, nothing uploaded unless you turn on a cloud model yourself.”

变现:免费基础版 + 一次性买断 Pro(29–49 美元终身授权,解锁说话人分离、PDF/Word 导出、多语言模型);或 BYO API Key 的云模型可选增值,主收入靠本地版买断,避免订阅疲劳。

MVP:2 周版:用 whisper.cpp + 全局热键做 Windows 本地听写,输出纯 Markdown 文件和文件夹结构,无账号无联网;第二周加上“麦克风与系统音频分轨录制”,说话人标记先用简单的双通道区分(我 vs 其他人),不做完整声纹聚类,先验证本地优先是否有人愿意付费。

竞品:Otter.ai、Granola、superwhisper、MacWhisper、Fathom、Windows 自带 Voice Access、whisper.cpp 自建方案

摘要:一位非专业开发者用 Claude 写代码,6 个月做出 Windows 本地优先的听写 + 会议纪要应用 A9 Voice(Rust + Tauri + Svelte,免费),所有音频和 Markdown 笔记只存在本机,声纹和分轨录音也都在本地完成,云模型默认关闭。这验证了“隐私本地化 AI 听写/纪要”在 Windows 侧确实存在未被满足的需求,但竞品密集(Otter、Granola、superwhisper 等),且帖中没有任何付费意愿或收入数据,付费转化仍需自行验证。