Making a training video with a talking avatar

Most guides over-explain. This is the part that actually moved the needle.

Step two, give it one job, not five. The moment you pile on asks, the output splits its attention and you get a polite nothing. Synthesia helps here because it needs no camera and ships in dozens of languages before lunch.

Check the output against one real example before you trust it. One check beats a hundred promises.

Name the file something you'll find later. Future you is not as clever as present you thinks.

The catch is that it's only as good as the example you feed it. A weak example gives a weak result, and that's on you.

I'd have kept doing it the hard way if someone hadn't shown me this. Now I can't unsee it.