Jeff For Local Marketing (YouTube) discusses recent improvements in AI avatars—especially lip-sync accuracy and emotional/intent interpretation—reporting that avatars now behave more like actors rather than robotic readers. He describes personal experiments feeding videos of himself into platforms (e.g., HeyGen, Synthesia), tweaking prompts to convey intent, producing multiple variants (including AI-voice/avatar), and observing convincing results (recipients unable to tell an avatar from real Jeff).
Chapters
When. When we saw that we were taking agents or videos of myself, if you will, feeding it back into a platform like hey Gen or Synthesia. And then it would lip sync almost perfectly. The audio would be almost perfect, if not even better. The one thing that we have to play with, and maybe it's me, I have to create some image or some videos without my hands moving around. That would probably do a little better job. But we can now play with all that because where before it was, it was robotic, it was just reading the script. Now those same avatars are like actors and actresses. It interprets the script and the intent. Now I'm playing around with this a little bit to actually add some of these words to help it understand the intent of. Of that particular line. Seriously, they said or with. With deep thought. I'm just playing around with that. I don't have a particular. I don't have any answers on that yet. But what I am seeing is the results for avatars for our clients are just getting better and better and better. In fact, I sent one out to a dear friend of mine. Of mine. I created three different videos and one of them was an AI voice agent, or excuse me, my avatar sent it out and she said, I showed it to my husband. We didn't know for sure, but when it was done, I told him that was an avatar that wasn't really Jeff. And he's like, no way. That's a cool.
