AI-Generated Avatars Come To Life

Aincent

March 18th, 2023

Using Generative AI To Create Talking Avatars

AI image avatars have hit mainstream adoption lately.

Lensa AI’s magic avatars have populated our social media feeds, and MyHeritages’ AI time machine is making the creation of selfie avatars fun.

But the next step in AI-generated avatars lies in the creation of videos that appear lifelike. While there are various apps and generative tools exploring this, the most impressive tool I’ve seen comes from D-ID’s Creative Reality Studio.

If you are unfamiliar with D-ID, they are the company behind the technology used in MyHeritage’s Deep Nostalgia which went viral on social media in 2021.

Last year, D-ID improved the technology by adding the ability to animate faces synchronized with audio in a new MyHeritage feature called Deep Story.

But D-ID has also released a self-service video platform called Creative Reality Studio that allows users to animate presentations from a single image.

So the idea is you could either use a predefined presenter, upload your own photo avatars, or generate new AI avatars directly in the platform via text prompts:

In addition to generating new visual AI presenters, D-ID also includes the option of using AI (GPT-3) to help you write your script…

So this is a full-package Generative AI tool that utilizes text-to-image avatar creation via Stable Diffusion, text creation using GPT-3, and combines that with D-ID’s deep-learning face animation technology.

So I decided to test this myself.

In my recent posts, I had been experimenting with some AI-generated psychedelic art with a blend of superhero avatars, so I decided to use a couple of those image avatars and animate them.

The results are impressive.

Here is a video of several of my avatars explaining Generative AI…

If you feel like the predefined voices aren’t as natural as you would like, you can always upload recorded audio and it will be synchronized with the avatars.