VocalScroll v2.0: On-Device Voice Scrolling
This post is a record of the v2.0 release. VocalScroll has since been rebuilt as v3, and the app described below is the current one.
With v2.0, VocalScroll became a speech-activated video teleprompter that runs its voice recognition on your device.
Teleprompter apps used to mean one of two compromises. You got timed scrolling that ignores whether you speak, pause or improvise, or you wired your microphone to cloud transcription that charged monthly fees and sent your private audio to distant servers.
VocalScroll v2.0 ran a voice recognition model inside your browser. Speech recognition ran on your device, and your scripts and recordings stayed there too.
What is new in version 2.0
We rebuilt the software around privacy, speed and ease of use. The main additions:
| Feature | What it does | Why it matters for creators |
|---|---|---|
| VocalScroll Speech Assistant | Follows your voice as you read. | The text scrolls at your natural speaking pace and pauses when you pause. |
| On-Device WASM Engine | Loads a lightweight Kaldi-based speech model of about 40 MB into browser memory the first time you use a language. | Speech recognition runs on your device instead of on a server. |
| Unified Session Recording | Captures your webcam video and microphone audio together. | Your takes are saved on your device, in Takes. |
| Microphone level meter | Shows a live level bar for your voice. | You can see whether your microphone is active and picking up sound. |
Designing for privacy
Your drafts, speeches, pre-release reviews and private journals are yours. VocalScroll keeps them on your device.
- Speech processing on your device: Speech parsing and the microphone level meter run in your browser, and your recordings stay on your device.
- Local cache: The speech model for a language downloads the first time you use it, then loads from your device cache.
- No accounts: There are no login screens and no email signups. Pirsch counts page views on the public pages without cookies. Open the page and start recording.
How to get started
To start your first session:
- Load a script: Write or paste your speech text into the Script panel in the Studio.
- Set up camera and mic: Pick your microphone under Audio in the Studio's Inspector, and switch on Camera for your webcam. Choose Behind text under Camera view to see yourself behind the script.
- Choose voice mode: Select Voice in the Voice/Paced control.
- Hit Start: Click Start, wait for the 3-2-1 countdown, and read your script. The prompter tracks your voice.
VocalScroll is free and runs in a modern browser on desktop, tablet or phone.