📡 Idea Radar 每天 8:00 / 18:00 推送 · 18:00 生成精选
reddit · SideProject · 60 分 ← 返回列表

I built an offline dictation app that runs entirely on your own GPU (Windows/macOS)

Hey! I got tired of cloud dictation tools sending every sentence I say to a server, so I built Whisper Dictate: hold a hotkey, speak, release - the transcription is pasted at your cursor. Everything runs locally on your own machine. Audio never leaves it. What it does: - English + German profiles with switchable hotkeys - On-device streaming models (NVIDIA Nemotron 0.6B) on your own GPU - CUDA on Windows, MPS on macOS - with CPU fallback; faster-whisper profiles also available - Auto-edits,

五维评分

付费意愿
60
需求真实度
50
个人可落地
85
竞争空间
30
新鲜度
70

AI 分析

痛点:云端听写工具把用户每一句话上传服务器,隐私敏感用户需要完全本地运行、可离线格式化的听写方案。

目标用户:重视隐私的开发者、写作者、记者、律师、医生及多语言办公用户,尤其是 Windows/macOS 桌面端高频听写人群。

证据:I got tired of cloud dictation tools sending every sentence I say to a server, so I built Whisper Dictate... Everything runs locally on your own machine. Audio never leaves it. Pricing: there is a free community edition, and the full Pro version is 40 EUR/year after a 14-day trial (no card needed for the trial). Source is public

变现:Freemium:免费社区版引流,Pro 版 40 EUR/年订阅;后续可加团队授权、企业私有化部署、行业热词包或格式化模板市场。

MVP:用 faster-whisper 或 NVIDIA Nemotron 0.6B 做本地流式转写,实现全局热键按住说话、松开后粘贴到光标处,先支持 Windows/macOS 单平台,提供基础标点/换行/删除词等确定性后处理,2 周内可出可试用版本。

竞品:Superwhisper、MacWhisper、Wispr Flow、Otter.ai、whisper.cpp、Whisper Dictate

摘要:独立开发者发布离线听写应用 Whisper Dictate,按住热键说话,松开后本地转写并粘贴到光标,支持 Windows/macOS、本地 GPU 和英德双语,免费社区版加 Pro 40 EUR/年,源码公开。它验证了隐私优先听写产品的付费模式,但缺少外部用户付费或评论数据,且同类竞品较多、源码公开导致壁垒偏低。