top of page

How to Create Video Assistants/Avatars with AI

As AI image/video generation tools become more ubiquitous, some faculty (and college presidents) are experimenting with these tools to create AI avatars.

If you are interested in creating an AI avatar, watch this video (NYU NetID required) narrated using a clone of Dr. Kristine Rodriguez Kerr's voice (with her permission).


You can also follow the steps below:


First, make an AI clone version of your own voice. This will allow you to use text-to-speech using your cloned AI voice so you do not have to physically record any audio/video. To clone your voice, create an account on Play.HT and upload a 30 second mp3/audio file of you speaking clearly. We recommend recording yourself using a smartphone or any external microphone while reading from this "Training Script." The Cloning process will take only a few moments and you can start creating text-to-speech files once it is done cloning. Play.HT's free account has a 5000 word limit and we recommend breaking up the script into sentences so if the cloned voice isn't accurate, you can "regenerate" only that specific portion of the script rather than the entire script. Once you are satisfied with the script, click on the "Export" drop-down menu and select "As a single audio file."


The second step is to upload this audio file and a headshot to Studio D-ID. Studio D-ID is an AI generative service that can animate 2d images of people's face and have the face and lips sync'd up with text/audio. A free account on Studio D-ID will allow you to create a 2 minute video (additional credits each day for up to 5 total minutes at the end of the free trial). Upload a headshot of using their example and follow the instructions provided to ensure the animation is accurate:

  • Medium shot, frontal facing

  • Neutral facial expression, closed mouth

  • Minimal head size within the image - 200X200 pixels

  • Good and solid lighting

  • Size - up to 10Mb

  • No face occlusions

You will then upload the downloaded .wav file from Play.HT to Studio D-ID. Once the audio file has been uploaded, click on the "Generate Video" button and wait a few minutes for the animated AI video to be generated. The video can then be downloaded as an MP4 file for future use. Please note that there will be a faint watermark over the video for trial users and a minimum of $5/month subscription will have the watermark be removed.


References

Rooney, P. (2023, October 10). College of Southern Nevada deploys AI avatar to better

engage students. CIO. [LINK]

299 views

Comments


bottom of page