When NOT to Use a Talking Avatar (3 Cases Where It Backfires)

I'm pretty bullish on talking avatars. I post one most days. But there are a few situations where reaching for the avatar tool actively hurts the content. Here are the three I've learned the hard way.

1. When you need to react to something happening in the world

Breaking news. A conference moment. A tweet that's blowing up. The window for reactive content is hours, not days. By the time you write a script, render it, and post it, the moment is over.

More importantly, reactive content has emotional weight that a stylized avatar can't carry. When something genuinely surprises you, the audience wants to see you processing it in real time — not a cartoon version of you delivering a curated script about it. The shrug, the "what?", the half-laugh — those don't translate through an avatar render.

For reactive content, take the phone out, hit record, post raw. The avatar workflow is for the planned content that benefits from polish. Reactive isn't that.

2. When the script depends on a physical environment

"Look at this." "Watch what happens when I do this." "This is what I built today." Anything that requires the camera to show something other than a talking head doesn't work in HeyGen.

You can fake it with a background swap, but the audience can tell. The talking avatar standing in front of a still image of "the thing" reads as detached. The energy of "I'm holding the thing, watch me show it to you" is what makes show-and-tell content work, and avatars can't reproduce that.

For show-and-tell content, you're back on camera. Or you're producing it as a different format entirely (a slideshow, a screen recording, a static image). The avatar is the wrong tool.

3. When the script needs to feel intimate or vulnerable

Sometimes you want to share something personal — a hard decision, a setback, a real reflection. The avatar can deliver the words, but the words land flat because the audience knows the character isn't actually feeling anything. Vulnerability has to feel risky to land. A polished render of a vulnerable script feels like a Hallmark card.

If the content is truly personal, take the phone out and record it shaky. The shake is the signal. Avatars are great at delivering low-stakes observation. They're terrible at delivering high-stakes emotion. Don't ask them to do the wrong job.

The frame

Talking avatars are for content with these traits: planned, observational, low-to-medium stakes, doesn't depend on physical objects in frame, audience already knows the character. That's most of what creators post. But it's not all of it.

The mature use of HeyGen is knowing what's not a HeyGen post. Save the avatar for the leverage situations. Use the camera for everything else. The two formats together cover everything; either one alone leaves real gaps.

— Jeff