ShareWhen you work at a digital marketing agency, innovation is the name of the game. At Elixirr Digital, we’re always exploring new ways to elevate our content and connect with our audience on a deeper level.
An idea that recently sparked my interest was incorporating videos into our blog posts—not just as an eye-catching addition, but as an essential tool for increasing engagement and providing accessible content for all users.
I don’t have time to sit and record a video for every blog post I write, even though that would be the perfect solution. I also have a face made for radio, so spending time in front of a camera isn’t my ideal.
Then it hit me: Could I use AI to solve both problems? Let’s dive in and find out…
The challenge: Time-consuming video creation
We all know video is a powerful medium, but let’s be real—it’s also incredibly time-consuming. Crafting a polished video can take hours, from scripting to filming to editing. That’s time we often don’t have. So, I set out on a mission to find a more efficient solution that still delivered on quality.
ElevenLabs
My first experiment was with a tool called ElevenLabs, which takes text content and turns it into speech, using a range of preformatted options. I gave it the text from one of my recent blog posts and it went to work. The output, even from the free version I used, was pretty good but left me with an audio file, so I would still need to spend time creating some kind of video to go alongside it.
Invideo AI
My second experiment involved full AI video creation using a tool called Invideo AI, which promised to take my text and turn it into a full video, including a voiceover, subtitles, background video clips, background music and transitions to boot! Sounds perfect, right?
I fully hold my hands up and admit that there could be more to the process than meets the eye, but having fed my blog post text into Invideo (which I thought might do most of the work for it) it pretty much entirely rewrote the script and turned it into a new piece of content entirely. Great if you’re creating explainer videos, not so great if the video is supposed to accompany an existing piece of content. Two groups of users getting different information depending on their choice of content? No, thank you! The output was very well done though, which is impressive given that I was again using a free version of the tool.
For some reason, it also decided that the narrator should be an AI version of David Attenborough, which while very pleasant and familiar, felt a little bit weird given that he has no affiliation with us.
So, we need a tool that follows a script exactly and outputs a video file with minimum input.
Enter HeyGen—a tool that promises to create instant avatars and scripted videos with minimal effort. I decided to put it to the test and see if this technology could meet our needs.
The experiment: Creating my digital double
The process began with creating an instant avatar. A digital version of myself, which could be used over and over again. HeyGen’s platform made it surprisingly easy. I recorded a 30 second video using my webcam, and within minutes, I had a digital version of myself ready to go.
Next, I added a script, and HeyGen generated a video where my avatar delivered the content. It was as simple as it gets—no camera, no lights, just a few clicks.
The result? Pretty impressive, but not without its quirks.
The modelled voice resembled mine, but it wasn’t a perfect match. There was something slightly off about it—it was as if someone had captured my voice on a particularly uninspired day. The intonation was flat in places, making the delivery feel somewhat robotic. Then there were the movements. My avatar looked a lot like me, but some of its gestures were just creepy. It’s hard to describe, but it triggered the uncanny valley.
The improvement: Refining the avatar
Not one to be easily discouraged, I decided to tweak the process. I recreated my avatar, this time being more expressive during the creation phase and ensuring I maintained eye contact with the camera at all times.
This definitely improved the overall look of the avatar—there was a noticeable boost in the naturalness of the movements. However, the voice? Still flat. The modelled voice didn’t quite capture the nuances of my speech, leaving the delivery feeling somewhat lifeless.
It’s possible that by providing a longer, higher quality, recording of myself, I may be able to get better results, but that’s a question for a follow up blog.
The verdict so far: Promising, but we’re not quite there
HeyGen has a lot of potential. The ease with which I could create a digital avatar and produce a video was impressive, and for certain applications, even the trial version I used could be a game-changer. However, when it comes to fully replacing human-driven video content—especially for something as personal as a blog—it’s clear that more work is needed.
In the meantime, we’ll continue experimenting, because standing still isn’t an option. There is a world of engagement and accessibility just waiting to be unlocked and AI video generation could be the next big thing. Who knows? We’ll keep pushing the boundaries and exploring what’s possible. Watch this space!
If you think you could do more to make your website accessible, why not get in touch with our team of experts? We’ll be able to advise you on where to start and help guide you through the process of enhancing your own content.
