The Pipeline That Shouldn't Work This Well

Google Slides to Synthesia breakdown to avatar script. Three steps. Maybe fifteen minutes. My knowledge course pipeline just became something I don't fully understand yet.

Here's the chain: I built the slides like normal. Showed them to Claude, who decomposed each slide into Synthesia-ready segments — timing, structure, what the avatar needs to carry. Then showed that to Claude, who wrote the full talking-head script.

I didn't write a single word of the script.

The part that's hard to sit with is that this isn't some elaborate multi-tool setup or a custom workflow I spent a week engineering. It's just — show the thing, get the next thing, show that, get the final thing. The assembly line I would have spent an afternoon building is just a conversation.

The knowledge course format has always had this brutal bottleneck: you can know the material cold, you can have the slides done, and then you still have to translate visual information into natural spoken language, per slide, without sounding like you're reading. That translation work is where the momentum dies. That's the part that made courses feel expensive to produce even when the ideas were free.

That bottleneck is gone now. At least for me, today, in August 2024 — it's gone.

I'm not going to speculate about what this means for course creators broadly or the industry or whatever. I just made something faster than I've ever made it, and the quality didn't obviously suffer, and I need to go lie down.