You’ve probably seen these around, all AI avatars either teaching or showing off something on tiktok, youtube or your personal social video app of choice. I decided to go through the process of creating one myself just to understand the process. I’ll walk you through how I created it but check out the results below.
The first step was creating the character. To do this I used midjourney and simply described a portrait of a character that I wanted my avatar to look like. You can browse the midjourney community for inspiration.
Next I created a script. I wanted to do something like a Bob Ross style tutorial on designing user interfaces with artificial intelligence. So I went to ChatGPT and simply asked for one. The results were pretty good.
Next was the voice. There are many AI text to voice services, but to be honest most aren’t very good. I ended up using play.ht because they have some Ultra Realistic Voices which are the best I’ve heard. I used the free trial as it’s not cheap to use.
Now for turning the audio and image into a talking video. For this I used D-id which is incredibly simple. You just add your own image for a character and upload your audio. I made sure to replace the background of my character with a green screen so I can key out the video later.
I was pretty happy with the results and may play with this stuff a bit more in the future.