SushiSpeak

A native macOS app for English speaking practice. Set a countdown timer, hit Start — it records your voice, shows a live waveform, and saves the audio when time's up.

Features

• Countdown timer — MM:SS display, stepper controls for minutes and seconds, persists across launches
• Auto record — starts recording on Start, stops and saves on timeout or manual Stop
• Live waveform — animated bars driven by real microphone RMS level (flat when silent)
• Audio format — choose MP3 / M4A / WAV per session; format badge shown on each recording
• Gain boost — 2.5× input amplification so normal speaking volume comes out clearly
• Recording list
  • ▶ Play back inline
  • 📁 Reveal in Finder
  • 🎙 Transcribe with Whisper → result shown in popup, one-click copy
  • 🗑 Delete with confirmation; Delete All
• Whisper transcription — local on-device AI (whisper-cpp), supports Chinese / English / mixed
  • Models: Tiny / Base / Small (default) / Medium / Large V3
  • Download models in-app; progress shown in header
• System alert — sound + Dock bounce when session ends (no notification permissions needed)

Requirements

• macOS 13 or later
• Xcode Command Line Tools: xcode-select --install
• ffmpeg (Homebrew): brew install ffmpeg (for dev mode only — bundled in production build)
• whisper-cpp (Homebrew): brew install whisper-cpp (for dev mode only — bundled in production build)

Build & Run

cd ~/code/SushiSpeak
./build.sh          # release build: bundles ffmpeg + whisper-cli, opens app
./build.sh -d       # dev build: uses system Homebrew tools, faster iteration

Output: .build/SushiSpeak.app

Install permanently:

cp -r .build/SushiSpeak.app ~/Applications/

Whisper Models

Click 🌐 in the header to open the model download page, download any ggml-*.bin file, then click 📁 to select it. The loaded model is shown as a chip — click × to remove.

Download page: https://huggingface.co/ggerganov/whisper.cpp/tree/main

Model	File	Size
Tiny	ggml-tiny.bin	~75 MB
Base	ggml-base.bin	~142 MB
Small	ggml-small.bin	~466 MB
Medium	ggml-medium.bin	~1.5 GB
Large V3	ggml-large-v3.bin	~3.1 GB
Large V3 Turbo	ggml-large-v3-turbo.bin	~1.6 GB

Models are stored in ~/Library/Application Support/SushiSpeak/models/.

Recordings Location

~/Documents/SushiSpeak/rec_<timestamp>.<format>

---

版权声明

Copyright © 2026 EddieChan1993. All rights reserved.

本软件及其源代码受版权法保护。

• 禁止未经授权的商业使用：未获得作者书面授权，不得将本软件或其任何衍生版本用于任何商业目的，包括但不限于销售、出租、捆绑销售或以盈利为目的的分发。
• 个人学习使用：仅允许在获得授权的设备上用于个人非商业用途。
• 禁止二次分发：未经授权不得以任何形式转发、再分发本软件的安装包或源代码。

如需商业授权或合作，请联系：wx DC_Wen 或邮箱 dc_wen666666@163.com