Everyone's Using VEO 3.1 Wrong (Here's What's Actually Possible)

Cool, now that my controversial title has gotten your attention (don't ban me, pretty pls lol):

Some of these might look a bit dodgy because I made them whilst testing out automations and trying to refine the process, but basically I could automate a huge chunk of it when I was using n8n.

So you will also notice a lot of that stuff with the text, because number one, it's in Spanish. So I just quickly translated the dialogue so I could show the video to you guys, the one with the man wearing glasses.

And number two, since this was part of some testing I did with an automation. You may find repetitive things in terms of the actions of the character. Since I moved on from n8n, which is what I was experimenting on, I didn't polish the process.

It starts with creating the script, then adapting it to my own voice and style.

After that, you create the image prompts, use an API to generate the images, which become your starting frames/end frames, and then, based on those frames and the script, you create the JSON prompts.

This used to take me anywhere from three to five hours to create, but once I'd tweaked and refined it a bit, pretty much once I had my spreadsheet ready, I just copied and pasted the prompts into VEO 3.

It took me about 20 minutes from choosing the right clips to stitching them together in CapCut with auto captions.

Very basic editing since I'm extremely ignorant in that department.

I reckon this is the sort of thing that'll separate people who just churn out random slop from those actually creating quality content, the kind of stuff that could even help you land clients.

Still, I've seen even wilder stuff, so I need to up my game for my image and video prompting... not to mention the editing and music department.

2:03

1:57

1 comment