-
Using AI to Make video ads – the workflow that worked for Me on BF
The client lost their creative director in Nov. Before they left they delivered some baked assets – music, VO, and titles already on the videos, in weird lengths that couldn't be re-cut and repurposed for things like YouTube or CTV.
I did end up getting a dump of probably 100 hours of b-roll but without a creative team there was no hope of getting ads in market before BF.
This is the AI workflow I used to make ads that are doing surprisingly well.
-
Write the voiceover using ChatGPT. I found that as of now, ChatGPT is better for timing the V/O to the amount of seconds you need.
-
Convert that to a usable V/O. Elevenlabs is what is the norm for that but I ended up using Google's tool in Google's AI studio. I turned the temperature all the way up until it gave me an emotive reading.
-
Make a music track. I used Suno to make something original. The prompt described the emotions I was going after and it made something that worked.
-
Image assets – I like Google Flow. It uses Nano Banana and there are no watermarks.
-
Convert some image assets to video – Also Google flow. Many of the vides have flaws… but if you iterate enough you can get some very usable stuff.
-
Recompose existing images from 9:16 to 16:9 for end screens. Nano-banana is great at this with minimal prompting.
-
Edit it. Since this was mostly image assets we used plenty of Ken-Burns zooms. I like Davinci Resolve but there is a steep learning curve. I probably understand 2% of the platform.
The client crushed it so far. Video was only a small part of the execution but overall they way up Y/Y.
There are probably dozens of other ways to do this right now. Let me know if there was a better path that worked for you.
-
Log in to reply.