The pitch is great, which should be your first warning.
You type a sentence. A person who does not exist shows up on screen, in a blazer, in good light, mouth moving in roughly the right shape for the words you wrote thirty seconds ago. HeyGen, Synthesia, the Veo demos, the Sora clips everyone keeps reposting — there’s a moment with each one where you watch the output and go huh. Not huh, this is good. The tells are all over it. The eyes are off. The hands do a little loop. But huh, something just moved that I didn’t know could move.
I’ve been spending evenings in these tools the way I used to spend evenings in Photoshop around 2003, which is to say badly, and with the creeping sense that the software is teaching me more than I’m teaching it.
The thing I keep coming back to is that the avatars crossing the uncanny valley isn’t really the interesting part. They’ll cross it on some random Tuesday and nobody will send out a press release. The interesting part is what happens to your gut when video stops being the format you trust by default. We’ve had about a century of training that says the camera doesn’t lie, or at least lies less than the typist. That training has roughly the shelf life of a banana now, and as far as I can tell nobody — not the platforms, not the regulators, not the people building the tools — has any real plan for what comes next.
Which is, admittedly, a pretty grand thing to say at the bottom of a post about my new toy.
I’m going to keep playing with the toy.
/