
macOS的私密设备会议转录和语音输入。本地STT、摘要和操作项、完整导出、为AI助手提供MCP服务器。没有云端,没有API密钥。
概览
macOS私密语音转文字功能,永远不会离开你的Mac。 可以口述到任何应用,转录会议内容,让自己的声音和别人分开,并在设备上获取决策和行动项目的摘要。没有云端,没有账户,没有API密钥。 下载 ·会议 ·语音输入 ·需求 ·语音模型 ·从源代码构建 系统音频采集与实时文字记录分离,无需安装虚拟音频设备。Kyutai、Whisper和Parakeet都支持独立的会议通道。录音会自动处理:它会通知你日历会议开始(今天的活动也会在主页有开始按钮),在通知你会议结束时自动停止;Mac休眠时干净地关闭并提示唤醒时恢复;如果发动机熄火或麦克风消失,还能恢复或挽救会话。如果麦克风停止响应,横幅会提示“重启舒芙蕾”。 - 附有可编辑的笔记和日历参与者,旁边是。
概述(AI 翻译)
舒芙蕾
macOS私密语音转文字功能,永远不会离开你的Mac。
输入任何应用,转录会议内容,区分你的声音和别人的声音,然后得到 设备内的摘要,包含决策和行动项目。没有云端,没有账户,没有API密钥。
下载 · 会议 · 听写· 要求 · 语音模型· 源代码构建
知道谁在说话的会议
系统音频捕捉在实时文字记录中将我和他们**分开,无需安装虚拟音频设备。Kyutai、Whisper和Parakeet都支持独立的会议通道。录音会自动处理:它会通知你日历会议开始(今天的活动也会在主页有开始按钮),在通知你会议结束时自动停止;Mac休眠时干净地关闭并提示唤醒时继续;如果发动机熄火或麦克风消失,还会恢复或挽救会议。如果麦克风停止响应,横幅会提示重启舒芙蕾*。
- 实时文字稿,附有可编辑的注释,旁边显示日历参与者。点击单词可添加词典别名。
本页由 AI 翻译,技术细节请以原文为准。
README
以下为原文 README 节选,完整内容请查看原文。
Meetings that know who is talking
System-audio capture separates Me from Them in the live transcript, with no virtual audio device to install. Kyutai, Whisper and Parakeet all support separate meeting lanes. The recording looks after itself: it notifies you when a calendar meeting begins (today’s events also get a start button on Home), notices when the meeting seems over and stops on its own after warning you, closes cleanly when the Mac sleeps and offers to resume on wake, and recovers or salvages the session if the engine stalls or the microphone disappears. If a microphone stops responding, a banner offers Restart Soufflé.
- Live transcript with editable notes, and the calendar’s participants, beside it. Click a word to add a dictionary alias for it.
- Resume recording on a stopped meeting adds another session to the same meeting.
- From a meeting’s page: copy the summary or the transcript, export it (Markdown, JSON, SRT/VTT, or the audio as
.ogg), or delete it. Summary templates are editable in Settings > AI. - Optional audio, kept as compact Opus files for 7 days, 30 days or until you delete them, replayable with click-to-seek from any line. A meeting with system audio is stored as stereo (left = you, right = the other participants) so the two sides can be separated later, for example with
ffmpeg -map_channel 0.0.0/0.0.1; the built-in player plays them mixed. Stereo costs about 1.5x the disk of a mono recording. - Corrections that stick: fix a misheard name by hand once and Soufflé keeps that spelling, in a custom dictionary you can also edit yourself.
安装
This server does not publish a one-line install command.
Open the repository installation guide 新手?先看 mcpskill.org 教程 →