In my latest video, I demonstrate how to sort rows randomly in Excel using a collection of test questions and answers that will be featured in the next update of my book, “Self-Guided SQL“. What’s really interesting about this video is the voiceover provided by the A.I. text-to-speech app from ElevenLabs.io.

Voiceovers have always been the most challenging part of video production for me, from working out how to narrate the videos to dealing with microphone issues and ambient noise. Some video creators seem to have a natural talent for narrating and demonstrating something on screen at the same time; I need to write a script and record it separately from the screen capture. Even then, it’s very difficult to control the background noise where I am and the quality of my own voice can be a little unpredictable.

ElevenLabs solves all of that and the free tier of their service provided more than enough credits to cover this video and enable me to play around with some of the options. Even the first two paid tiers at $5 and $22 / month are very affordable for someone who’s doing even a few projects a month. I had looked at ElevenLabs before but I don’t remember their prices being this reasonable.

The actual app is divided into two sections, the Playground and the Products. I generated the voiceover in the Text to Speech section of the Playground which is used for short-form output. The Playground also includes the Voice Changer with which you can upload a voice sample or record a new one and have it changed into one of the thousands of voices available. The Sound Effects section accepts text descriptions of the effects needed and will produce a selection of them to choose from.

After checking out a few of the voices, I selected an American, male voice named “Chris” that sounded like one that would inspire confidence when talking about Excel and plugged in the script that I’d typed up after recording the screen output for the video.

I wanted to insert some pauses between the paragraphs and a quick Google search clued me in to using the <break time=1.5s” /> tag shown in the screenshot above which worked really well. This let me generate a single, downloadable MP3 which I then converted to WAV with Audacity and then imported into Camtasia. Then, I was able to easily split the audio file on its pauses and position the individual segments to sync with the video.

After I uploaded the video to YouTube, I discovered an embarrassing single word error in the video and was able to quickly regenerate the specific paragraph in ElevenLabs, replace it in the video and re-upload the video in maybe 5 minutes. Another feature of the text to speech is that it’s nondeterministic, meaning that if you generate the same text more than once, the inflections and even the pronunciation of specific words might not be the same. So, if you don’t like the tone in one output, just try it again. After a few tries, it actually got my name right!

There’s a lot more to see at ElevenLabs but this service alone wowed me so much that I’ll be happy to sign up for one of the paid tiers and I look forward to creating more videos with it. It’s definitely one of the best A.I. tools that I’ve used to date.