Voice-following teleprompter
Speak naturally. Keep your place.
CandidCue follows your voice on your own computer and keeps your next line at eye level. Ad‑lib, pause or start a sentence over, and it waits for you.
Free and open sourceNot released yetA desktop app for macOS, Windows and Linux
An animation of CandidCue's stage following a voice: the words already spoken dim as the text moves to keep the next line beside two markers. When the speaker goes off script the text waits, and it carries on once they are back.
How it works
Write it. Present it. Read it.
No scroll speed to set and no operator to brief. Three steps, and the last one is just talking.
-
Write your script
Write it in plain Markdown, the way you’d say it. Headings divide it into sections, and a cue in square brackets puts a note on the glass that you see but never say.
-
Press Present
Teleprompter mode fills the display you pinned, mirrored for a beam splitter if your rig needs it. Reading window mode keeps it in a window beside your webcam.
-
Read it aloud
The words you’ve said dim and the line you’re reading stays between the markers, at whatever pace you speak. Ad-lib and the text waits where you left it.
Features
A prompter that waits for you
Everything in CandidCue serves one job: keep your next words in the reading band, whatever you say on the way there.
Follows your voice, not a clock
Ad-lib, pause or start a sentence over, and it waits for you rather than running ahead. A deterministic alignment engine, not a language model, decides where you are. Nothing on the glass is generated.
Built for glass
Present puts the stage on the display you pinned, mirrored for a beam splitter. It knows the Elgato Prompter and Prompter XL.
Scripts in Markdown
Cues, pauses and part breaks are plain text. Write a word one way and say it another.
Takes and parts
Present from the beginning, from the cursor or from any part. Practise is a rehearsal that reports nothing, and every take ends with a short summary.
Readable at a distance
Type is sized for how far you sit from it, not in pixels, and set in Atkinson Hyperlegible Next. Calibration presets cover a desk, a webcam and the Elgato rigs.
Your files, on your disk
Every script is a Markdown file you can open anywhere, with a revision history you can browse and restore. Export all data makes one backup of everything.
A closer look
From first draft to final take
Real screens from CandidCue: where your scripts live, how you write them, and what you see during a take.
Library
Your scripts are files on your disk
Every script is a plain Markdown file on your own computer, so you can open it in any text editor and keep it however you like.
- Revision history you can browse and restore
- Import .md and .txt files, or drag them in
- Export all your data, and restore from an export
Editor
Write in Markdown, read it as a document
Switch between Edit, Preview and Both. Headings make sections, and a cue like
[Cue: look at the lens]becomes an amber pill: a note on the glass that is never spoken or followed.- Edit, Preview, or both side by side
- An outline, find and replace, and autosave
- Pauses, beats and part breaks as plain text
Reading window
One screen is enough
No prompter rig needed. Reading window mode stays an ordinary window, so you can read from a laptop with the webcam just above the words.
- Paper or Dark appearance
- Text size, line width and spacing to suit you
- The same voice following as on the glass
Take controls
When it isn’t sure, it asks
If CandidCue can’t tell where you are, the text holds still and the Take controls show the nearby lines you might be on. It never jumps to a place you can’t see.
- Pick the line you’re on, or keep reading
- Set my place, Previous sentence and Next sentence
- Auto-scroll as a fallback, with [ and ] for pace
Privacy
Your voice stays here
Speech recognition and following run on your computer. CandidCue does not upload your microphone audio or your scripts.
No account or API key
There is nothing to sign up for and no key to paste in.
No analytics client or crash reporter
Logs are written on your computer and are not uploaded. Support bundles are created only when you ask for one, and go only where you send them.
One-time speech model download, about 128 MB
Voice following needs it. It downloads only when you choose to install it, from the Welcome screen or Settings › Speech, and is checked against a SHA-256 checksum built into the app.
Update checks you can turn off
CandidCue checks updates.candidcue.com for a newer version shortly after it opens and every six hours. You can turn that off in Settings › Data, privacy & updates. An update installs only when you choose, never during a take.
The network inventory lists every network path in the app and what each one sends.
Download
Free, for macOS, Windows and Linux
Free and open source under the MIT licence. Updates are signed and verified, and the app offers each one for you to install.
Not released yet
macOS
Universal .dmg, signed and notarised
Requirements
- macOS 12 or later
- Apple Silicon or Intel
The download appears here once it is published.
Not released yet
Windows
Installer (.exe)
Requirements
- Windows 11, 64-bit
- Windows 10 should work but is not tested
The download appears here once it is published.
Not released yet
Linux
.AppImage or .deb
Requirements
- 64-bit, Ubuntu 22.04 or newer
- WebKitGTK 4.1
The download appears here once it is published.
On every platform
- A microphone
- A second display for a prompter rig (optional)
- English, for voice following
The speech model
Voice following needs a one-time speech model download, about 128 MB, which you start from the Welcome screen or Settings.
Updates
CandidCue checks updates.candidcue.com for a newer version shortly after it opens and every six hours, unless you turn that off, and installs one only when you choose to, never during a take. Updates are signed and verified.
Questions
Before you record
Short answers on cost, rigs, privacy and installing.
Still have a question? Ask on GitHub
Yes. CandidCue is free and open source under the MIT licence, with no paid tier and no account to create. The code, the releases and the issue tracker are all on GitHub.
No. A laptop and its webcam are enough. CandidCue presents in one of two ways, and you can switch between them during a take:
- Reading window keeps the script in a window you can size and move, so it can sit just below your webcam.
- Teleprompter fills the display behind your rig’s glass, and can mirror the text for the beam splitter.
Yes, the Elgato Prompter and the Prompter XL both work. Choose the Prompter as your display in Settings › Prompter display and rigs. The calibration has a preset for each, which sizes the type for how far you sit and offers to mirror it for the glass.
Mirror the text in one place only. If Elgato Camera Hub already flips the Prompter’s picture, leave CandidCue’s Horizontal mirror off.
It holds. While you say something that isn’t in the script, the text stays where you left it, and it moves again once CandidCue finds your place. Start a sentence over and it follows you back. When it isn’t sure, it waits rather than guessing ahead.
To move it yourself, pick your line in the Take controls or choose Set my place. If it keeps losing you, press A for auto-scroll. The docs list every take shortcut.
Only to download the speech model and to check for updates. Speech recognition and following run on your computer. Once the model is installed, you can write and present without a connection.
The speech model is downloaded when you choose to install it. It is about 128 MB, and you fetch it once. You can turn the update check off, and the network inventory lists every network path in the app.
No. CandidCue does not upload your microphone audio or your scripts. Speech recognition and following run on your computer. Your scripts are Markdown files on your own disk.
Support bundles are created only when you ask for one, and go only where you send them. The privacy policy has the detail.
Yes. The speech model CandidCue installs is an English one, so voice following works with English speech only.
Scripts in other languages still work. Press A for auto-scroll during a take, or move the script by hand.
Any microphone your computer sees: built in, USB or Bluetooth. The closer it sits to you, the less of the room it hears. Settings › Audio has a level meter and Test transcription, so you can check it before a take.
The installer isn’t code-signed yet, so Microsoft Defender SmartScreen stops it with a “Windows protected your PC” notice. Choose More info, then Run anyway.
To check the file first, run Get-FileHash on it in PowerShell and compare the
result with its line in SHA256SUMS-windows.txt on the release page.
CandidCue checks updates.candidcue.com for a newer version shortly after it opens and every six hours, unless you turn that off, and installs one only when you choose to, never during a take.
When there is one, a notice in the corner of the window offers Install and Restart or Later. Installing restarts CandidCue, and your scripts and settings stay as they are. A Linux .deb can’t replace itself, so there the notice offers Download instead.
Updates are signed and verified. To stop the check, turn off Check for updates automatically in Settings › Data, privacy & updates.