Speechius's answer

It recognises the words you actually say and finds them in your script, rather than scrolling on a timer or reacting to how loudly and quickly you speak. That means it still follows you when you rephrase a sentence, skip a line or jump back.

Recognition runs entirely on your own device, and the window can be excluded from screen sharing and screen recording.

Speechius's answer

Several teleprompters advertise voice-controlled scrolling, but most of them track your speaking pace and volume: they speed up when you speak faster and stop when you go quiet. They do not know which words you said, so going off-script or jumping back confuses them.

Speechius matches the recognised words against your script, which is the part that keeps working when you stop reading and start talking. It is also a one-time purchase rather than a subscription, and nothing is sent to a server.

Speechius's answer

People who speak to a camera or an audience while their script has to stay out of sight:

• Creators recording screencasts and tutorials
• Trainers and teachers running online classes and webinars
• Speakers rehearsing conference talks

Typically solo, on their own machine, without a studio setup.

Speechius's answer

I record programming tutorials for my own YouTube channel, and neither option worked for me: reading from a script sounds read, and improvising means losing the thread until a ten-minute recording takes forty. Existing teleprompters scroll at a fixed pace, so I ended up chasing the text instead of explaining.

I built the thing I wanted, and it turned into my first own product, released in July 2026.

Speechius's answer

• Flutter for the user interface on macOS and Windows
• Apple Speech for on-device recognition and window exclusion on macOS
• Vosk, bundled as a local model, for recognition on Windows
• Ed25519 signatures for licence keys, sold through Paddle
• Sparkle for updates on macOS, Sentry for crash reporting