Launching today

Sovereign Voice

Launching today

Your voice, on your silicon. No cloud required.

3 followers

Your voice, on your silicon. No cloud required.

3 followers

Visit website

Runs Qwen3-TTS 0.6B entirely on your Mac via MLX. No cloud, no API keys, no per-character fees. Voice cloning from a 3s sample, 10+ languages, ~1s latency on M-series chips. Free download with 25 generations. Pro is $19 one-time for unlimited generation.

Free Options

Launch tags:Productivity•Artificial Intelligence

Launch Team

Fin Startups get Fin free for a year + 93% off Intercom

Promoted

Maker

📌

Hey Product Hunt, I built Sovereign Voice because I was paying a cloud voice service to store and synthesize my own voice, and one day it occurred to me: my MacBook Pro has a dedicated neural accelerator sitting idle most of the time. Why is my voice living on a server I do not control? The problem is not that cloud voice services are bad. It is that their architecture is wrong for a certain kind of user. If you are a podcaster, a voiceover artist, or someone who records sensitive material — interviews, legal narration, confidential briefings — you have two options today: pay per character and trust a third party with your voice model, or use nothing. Sovereign Voice is a third option. The architecture is straightforward. You download the Mac app. It installs a daemon that runs the Qwen3-TTS voice synthesis model via MLX on the Metal GPU. The daemon lives in your menu bar. When you want to synthesize, you open the studio window, type your script, click Synthesize. The audio is on your drive in 5-9 seconds on an M4. Pull the ethernet cable. It still works. A few things I am proud of: - The full synthesis pipeline runs without an internet connection. - There is no account. Nothing to sign in to. - The voice model file lives on your disk. You own it. Back it up. Move it to a new Mac. - Opt-in anonymous telemetry only — no audio, text, or personal data ever sent. The Mac app is free with 25 generations. Pro is $19 one-time (unlimited generations, all voices, studio mode, custom clones). Drop your email after your second synthesis and get +5 free generations — no purchase required. Happy to answer questions in the comments.

Report

2d ago

Maker

A few people have asked about performance — here are the real numbers from our testing:

• M4 MacBook Pro: 5-9s for a 100-word synthesis
• M3 Pro: 8-12s for the same
• M2 Air: 12-18s
• M1 (base): 15-25s

The model is Qwen3-TTS 0.6B running in 8-bit quantization via MLX on the Metal GPU. It's ~1.9GB on disk. First launch downloads the model; after that everything is offline.

If you're on an Intel Mac, it won't work — MLX requires Apple Silicon. Sorry about that.

Happy to help anyone who's trying to get it running. Drop a comment or email contact@eidetic.works.

Report

4h ago

Maker

Demo video is up — watch zero-shot voice cloning of Trump, Freddie Mercury, and Morgan Freeman running entirely on-device via Qwen3-TTS + MLX. No cloud round-trip, no API calls. The whole pipeline runs on the Metal GPU.

Watch it here: https://pub-d695698bdc934fa19593...

All three voices were cloned from 10-second reference clips. WER (Word Error Rate) was 0.000 for Trump and Freeman at 10s — perfect transcription. The app ships with this capability built in.

Download the app: https://sovereign-voice-landing....

Report

3m ago