

QwenVoice
QwenVoice is a native SwiftUI macOS application that brings state-of-the-art text-to-speech to Apple Silicon Macs with no Python install, no terminal, and no dependencies required of the user — just download and run.
Features
Properties
- Privacy focused
Features
- No registration required
- AI Voice Cloning
- No Tracking
- Command line interface
- Works Offline
- Ad-free
- Text to Speech
- Waveform
- Voice synthesis
- Support for Keyboard Shortcuts
- Apple Silicon support
QwenVoice News & Activities
Recent activities
POX added QwenVoice as alternative to ElevenLabs, X to Voice, Voice Engine and FakeYou- POX added QwenVoice
QwenVoice information
What is QwenVoice?
QwenVoice is a native SwiftUI macOS application that brings state-of-the-art text-to-speech to Apple Silicon Macs with no Python install, no terminal, and no dependencies required of the user — just download and run.
It runs the Qwen3-TTS model family entirely offline via Apple's MLX framework, delivering fast, low-latency, low-heat inference on M-series chips. The app communicates with a Python backend over JSON-RPC 2.0 via stdin/stdout, managed transparently as a background process.
Features:
Custom Voice & Voice Design
Generate speech using 4 built-in English speakers (Ryan, Aiden, Serena, Vivian) or create entirely new voice identities from a text description (e.g. "deep narrator", "excited child"). Both modes are controlled entirely through natural language instructions — there are no sliders or SSML tags. The underlying discrete multi-codebook language model natively interprets prompts to modulate breath, pitch, resonance, and emotional delivery.
Voice Cloning
Clone any voice from a short 5–10 second audio sample (WAV, MP3, AIFF, M4A, FLAC, or OGG). Optionally provide a transcript of the reference audio to improve accuracy.
Model Manager
Download and manage MLX models directly from HuggingFace inside the app. No browser or command line needed. Uses a native URLSession-based downloader with real-time progress tracking.
Generation History
Every generation is persisted to a local SQLite database (via GRDB). The History view lists generations sorted by date (newest first) and supports text search filtering. Each entry can be played back instantly, revealed in Finder, or deleted.
Batch Generation
Submit multiple text entries for sequential generation in a single session.
Additional features:
- Temperature & max-token controls — Fine-tune the model's sampling behaviour from the UI
- Waveform visualisation — Live waveform rendered for generated audio clips (via AVFoundation + vDSP)
- Reveal in Finder — Jump directly to any generated file (Cmd+Shift+R)
- Keyboard shortcuts — Cmd+Return to generate, Space to play/pause, Cmd+. to stop, Cmd+Shift+O to open the output folder
- CLI companion — A standalone Python CLI in cli/ for headless or scripted use





