Building the BEST Video Resume Tool in 5 days — DAY 2
Also, check out Day 1: The First Steps
📅 DAY 2: The Teleprompter
I knew that just generating a script and letting users record themselves on webcam wasn’t enough.
I wanted to build a tool that actually makes it comfortable and extremely easy for people to record themselves.
First, I added a simple teleprompter with auto-scrolling text. That was the obvious move.
But then I thought… what if I could actually recognize the words you’re saying while you speak?
The teleprompter could scroll at your pace, and highlight exactly where you are.
I started by checking word by word, but it felt a bit clunky and unreliable.
Tiny pronunciation differences would throw it off, especially if you’re not speaking in your native language (like me).
If a feature like this only half-works, it’s worse than not having it at all. It needs to be reliable, otherwise, it adds an extra level of frustration: sometimes when you mess up as part of the normal process, but even worse when you were doing everything right and the teleprompter lagged, forcing you to start over.
So I scrapped the word-by-word detection and switched to sentence detection, and that made all the difference.
There’s a bit of wizardry going on under the hood, and now it feels way more stable and reliable.
The cherry on top is that the recording stops soon after it detects the last sentence. So you don’t get that awkward pause where you reach for the mouse to click stop. It just ends on its own, nice and smooth.
I also made it so that clicking anywhere or pressing any key will stop the recording too, just in case you want a bit more control.
For example, you could be holding the mouse and simply click to stop it without breaking your pose, since any click will end the recording.
One limitation: the speech recognition part only works on Android and desktop (both Windows and Mac). It doesn’t work on iOS, or at least I still can’t get it to work.
If you’re on iPhone or iPad, it’ll default to auto-scroll, which is still available for everyone.
All sounds good, right? But the truth is, none of this matters if the teleprompter still makes you look like you’re just reading off the screen.
You’re supposed to show your human side (or at least fake it elegantly) not read like a robot.
So here’s what I did: I placed the teleprompter at the top of the screen, centered right above the webcam.
That way, when you’re reading, it actually looks like you’re making eye contact (especially if you step back a little from the device).
I also kept the text narrow so you’re not scanning side to side, which would make it obvious you’re reading.
And of course, you’re not supposed to stare at the words the whole time.
I’ll probably add a small tip in the UI to remind people to break the stare from time to time and keep it human and natural.
Day 2, done! ✅
Wy more complicated than I thought, but huge progress.
Update: Check Out DAY 3