Vocal Canvas BETA
v0.1.1
All Projects source code (github)

Vocal Canvas

v0.1.1 (Beta)

Public beta: features, UI details, and packaging behavior may change quickly.

Install on macOS

  1. Go to the Download tab.
  2. Download VocalCanvas.dmg.
  3. Open the DMG and drag Vocal Canvas.app to Applications.
  4. Open the app from Applications (use right-click → Open on first launch if prompted).

Install on Windows

  1. Go to the Download tab.
  2. Download VocalCanvasWindows.zip and extract it.
  3. Install Python 3.10+ if needed.
  4. Run:
pip install -r requirements.txt
python vocal_canvas_windows.py

Features

  • Multi-page website with Home, Demo, Q&A, and Download tabs
  • Quick online TTS demo preview
  • Desktop app with full controls (voice selection, speed, generation, export)
  • Downloadable macOS and Windows builds
  • Unified dark visual style for site and app

What to Expect in Beta

  • Frequent design and feature updates
  • Potential installer and first-launch edge cases
  • Some sections still being expanded (including Q&A docs)
  • Rapid bug-fix releases between versions

V3TS

What is V3TS?

V3TS (Voice to Text to Speech) is a real-time voice mirror built into the Vocal Canvas desktop app. It listens through your microphone, transcribes what you say using an on-device Whisper model, and instantly reads it back through the selected system voice — all without any cloud round-trips.

On-Device Transcription

Transcription runs entirely on your machine using faster-whisper, a local implementation of OpenAI's Whisper model. No audio is ever sent to a server. The tiny model delivers around 230ms inference time on a 1-second audio clip, keeping the experience near-instant even when you're talking fast.

Feedback Loop Prevention

V3TS automatically mutes the microphone listener while the synthesized voice is playing, plus a 500ms tail to catch acoustic echo. This prevents the app from hearing and re-reading its own TTS output in a loop.

Virtual Microphone Output

When BlackHole 2ch is installed, V3TS routes the synthesized voice to a virtual microphone visible in macOS System Settings. Any other app — Discord, Zoom, OBS, and more — can select that virtual mic as its input and receive the synthesized voice directly.

Configurable Controls

Pick any macOS system voice, set the speaking rate, choose your input microphone, and select the Whisper model size (tiny, base, or small) to trade speed for accuracy. All settings are applied live with no restart required.

Feedback helps shape the beta release. source code (github)