Showrunner — EP-012 Package

Title

3 AI-proposed options · option 1 locked

HeyGen avatars are getting scary good Direct claim with the curiosity gap intact — and it matches what people actually search.

I cloned myself with HeyGen — clients couldn't tell First-person stunt framing — stronger hook, weaker search match.

AI avatars in 2026: the uncanny valley is closing Category framing — safest of the three, and the flattest.

Description & chapters

drafted by AI · edited by Deven

Description

HeyGen's 2026 avatar models crossed a line we didn't expect: in a blind test, our own clients couldn't reliably tell the clone from the camera footage. In this deep-dive we build a full avatar clone, put it in front of real client work, and break down exactly where the illusion holds — and where it still cracks.

You'll see what actually changed in 2026 (it's the micro-expressions, not the lip-sync), the complete clone-test setup you can replicate, unfiltered client reactions, and the honest failure modes — hands, laughter, anything off-script. We close with a practical verdict for agencies: which deliverables an avatar can carry today, and which still need a human on camera. New builds and deep-dives every Thursday.

Chapters

00:00Cold open
01:12What changed in 2026
03:05The clone test
06:48Client reactions
09:30Where it breaks
11:02Verdict + CTA

Thumbnail

Queued

Candidate A — dark bg, big face, "SCARY GOOD" callout · selected

Candidate B — real vs AI split test

Produced as a graphic task — final pass claimable in the producer queue.

Publish checklist

4 of 7 done

Final cut approvedJul 10
Title lockedoption 1
Description & chaptersdrafted
Tags5 set
Thumbnail final passunclaimed — up for grabs in the queue
End screen & cardspending
Pinned comment draftpending

Complete the checklist — publishing is manual in v1; YouTube API upload comes later.

Publish target
Thu Jul 16 · in 3 days

After publish

Performance reporting lands with the YouTube API phase.