hehehai/AudioSmith
VOICE STUDIO ON MAC WITH MLX
Local app for speech synthesis, voice cloning, audiobook narration, dubbing and transcription on Apple Silicon. Works offline via MLX with no cloud or Python.
// KEY FEATURES
- TTS, voice cloning, M4B audiobooks, video dubbing
- Transcription with diarization, voice isolation, dictation
- 37 ready voices, fast and high-fidelity clone, voice design
- Runs offline on GPU via MLX, agent control through MCP
// DEEP DIVE: HOW IT WORKS
AudioSmith is a native macOS app written in Go: all models run locally on GPU via MLX. A single input box on Home determines the task from text, link or file — TTS, dubbing, transcription or audiobook. Results become projects that move between tools in one click. AI agents can drive everything through MCP.
No cloud: models download once and work offline. 37 voices (mostly English/Chinese) with filters for language, gender, age. Cloning offers fast and high-fidelity (48 kHz) modes, plus voice design. Export to WAV, M4A, MP3, FLAC, Opus with LUFS normalization and subtitles.
Requires Mac with Apple Silicon (M-series) and macOS 14+. Version 0.5.0-beta.1 is beta, signed with Developer ID, notarized by Apple, auto-updates. Expect rough edges; models may consume much VRAM. Python not required, ffmpeg optional for extra formats.
// USE CASES
Best for: content creators, podcasters, localizers on Mac with Apple Silicon
- Create M4B audiobook from EPUB with chapter and role separation
- Dub video into another language preserving timing and intonation
- Transcribe meeting with speaker separation and SRT export
New repositories every 30 minutes
REDDYX AI scans GitHub 24/7 and ships the best AI/ML/Web3 projects to Telegram.
Join on Telegram