I've written for camera. I've written for HeyGen. They look similar from a distance — both are short scripts in your voice. Up close, they're different jobs.
If you write a HeyGen script the way you'd write a camera script, the avatar will deliver it but the energy will be slightly off. Not bad. Just off. Five things change.
1. No filler "ums" or pauses
On camera, you can pause. You can let a beat hang. You can say "uh" and have it come across as authentic. In HeyGen, every word gets pronounced cleanly, which means filler words feel weirder than they would on camera. Cut them.
Free tier is more than enough to ship your first talking avatar — no card required.
Replace pauses with shorter sentences. Instead of "I was thinking… maybe… we could…" write "I was thinking. We could." The avatar reads the periods and gives you the pause for free, but cleaner.
2. Sentences should be longer than tweets, shorter than paragraphs
HeyGen handles ~10–15 word sentences beautifully. Anything shorter feels choppy because the avatar has to keep "resetting" between sentences. Anything longer and the lip sync starts to drift on the back half.
The sweet spot is two-clause sentences. "I almost cancelled this, would've been a mistake." "Thought it was overkill, turns out it's the one I use most." That structure suits both the avatar's pacing and the platforms you're posting to.
3. Numbers go in digits, not words
If you write "nine dollars," the avatar will say "nine dollars," but in transcription and accessibility readouts the number doesn't get the digit's weight. If you write "$9," you get "nine dollars" verbally and "$9" visually if any text appears. Always use digits. They hit harder.
4. Don't write the gestures
On camera, you might note "[shrug]" or "[looks at coffee]" in your script. In HeyGen, those notations either get spoken aloud or get awkwardly skipped. Either way, they don't translate. If you want a shrug-energy line, write the line so it carries shrug energy: "Whatever. Pinned now."
The avatar's micro-movements come from the script's rhythm, not from stage directions. Trust the rhythm.
5. End with a clear final beat
HeyGen renders the closing frame for a half-second after the last syllable. If your script trails off ("…and yeah, that's about it"), the closing frame holds on a trail-off, which reads weak. End on a clear beat: a short final sentence, ideally with a period that closes the loop.
"Sending it." "Pinned now." "That's the play." Three-word closes work great because the avatar lands the last syllable cleanly and the closing frame feels deliberate.
The shift in mindset
Camera scripts are about what you can naturally say. HeyGen scripts are about what looks deliberate when an avatar says it. The constraints sound limiting and end up making your scripts tighter. After a few weeks, you'll find yourself writing camera scripts in HeyGen style anyway. They're better that way.
— Jeff
Free tier is more than enough to ship your first talking avatar — no card required.