AudioSmith Open-Source Mac Voice Studio Adds Audiobook, Dubbing and Transcription Tools

AudioSmith, a free and open-source voice studio for Apple silicon Macs, turns text into speech, clones and designs voices, converts EPUB files into chaptered M4B audiobooks, dubs videos and transcribes recordings with speaker identification. All models run locally on the Mac's GPU via MLX, with no account, cloud service or Python required, and the app works offline once models are downloaded.
Event version 1.
Key insights
Local processing on Apple silicon
Every model runs on the Mac's GPU with MLX. The app requires no account, no cloud service and no Python, and works offline once its models are downloaded.
EPUB to chaptered M4B audiobooks
Importing an EPUB automatically splits chapters by heading and generates an M4B audiobook with chapters and cover art. Interrupted renders can resume, and editing a sentence regenerates only that sentence.
Voice design and cloning
Voices can be described in a sentence, such as a warm female voice around thirty like a documentary narrator, or cloned from a recording. The app includes 37 curated voices, mainly Chinese and English.
Video dubbing pipeline
Dubbing separates vocals from background audio, then transcribes, translates and dubs line by line using each speaker's cloned voice. Output duration stays close to the original video.
Transcription and dictation
AudioSmith transcribes recordings with speaker diarization and types dictated speech into any app.