reddit · SideProject · 10-07 05:50 73 分 ← 返回列表

I built an app that lets you talk to your computer instead of typing

Hi everyone, I'm Abhishek. Back in June I was paying for a dictation app. It turned my voice into text, and that was it. What I really wanted was something I could talk to: clean up what I said, help me reply to a message, take notes during a call. So I started building it on the side. 1.0 came out in July. Since then it's been downloaded over 3,000 times and picked up 260 stars on GitHub, and people sent me a lot of feedback. I spent the last few weeks reworking the whole a

五维评分

付费意愿

需求真实度

个人可落地

竞争空间

新鲜度

AI 分析

痛点：现有听写应用只把语音转成文字，无法理解自然口语中的犹豫和改口，也不能基于选中文本执行翻译、回复、解释等操作；用户想要一个本地隐私、可对话的语音助手。

目标用户：知识工作者、开发者、写作者、频繁口述或开会的用户，尤其是注重隐私的 Windows/Mac/Linux 桌面用户。

证据：Back in June I was paying for a dictation app. It turned my voice into text, and that was it. What I really wanted was something I could talk to... Since then it's been downloaded over 3,000 times and picked up 260 stars on GitHub

变现：开源核心 + 付费专业版/订阅：高级本地或云端模型、团队会议总结、跨设备同步、企业私有部署、优先支持；也可做一次性买断的桌面增强版。

MVP：2周MVP：基于本地 Whisper 做全局快捷键听写，用小模型或规则清理“嗯/啊”和改口，支持选中文本后调用 LLM 翻译/回复/解释，并直接输出到当前输入框；先只支持一个平台，不做会议记录和完整语音对话。

竞品：Wispr Flow、superwhisper、MacWhisper、Otter.ai、Granola、Willow Voice、Windows/macOS 内置听写

摘要：独立开发者 Abhishek 发布 SpeakoFlow 2.0，支持自然口述清理、选中文本操作、语音对话和会议纪要，默认本地运行且 MIT 开源。此前他付费使用听写应用，1.0 发布后下载超 3000 次、GitHub 260 星。这说明隐私优先的 AI 语音输入/助手有真实需求，但竞品多且原产品免费开源。