Today I tried to do some experiments with ai to create some short clips. Honestly, it was all a bit ad hoc and hopefully by typing out what I’ve done will help me understand what is working and what isn’t working with these experiments.
I wanted to create a series of clips with the same subject matter, but edited in different genres.
First I looked at different platform options, including runway (couldn’t get a subscription to work), adobe firefly (didn’t want to pay for it), Davinci (1.99 for a week), Google ai studio (didn’t want to pay), Canva (didn’t want to pay), synethesia.io.
Prompt: ‘I want a video of a woman sat in a modern kitchen at a table. I want her to pick up a glass of water and drink it. I want it repeated in the following styles:
Action, Documentary, Sci-fi, Horror, Noir, Romance’
Google gemini: prompted individual genres. 3.5 flash (two videos generated)
Synethesia: I want a video of a woman sat in a modern kitchen at a table. I want her to pick up a glass of water and drink it. Styled in a range of cinematic genres
also prompted I want a video of a woman sat in a modern kitchen at a table. I want her to pick up a glass of water and drink it. styled like a horror film
davinci: seeddance 2.0 two videos using the original prompt
I also tried using claude code to create a script using ffmpeg to edit video using scene detection and rules for the cuts. It didn’t work well: it needed existing cuts to make decisions about how it was going to change the scenes. I filmed a shitty video of my flat on my phone, and so the programmatic cuts didn’t make any sense.
I’m also desperately hot in my flat, its absolutely sweltering.
Note to add in my supervision meeting notes that Rachel offered to set me up with an editing place.
Then I’ve fed screenshots of the davinci ones (the best ones in my opinion) and asked it to guess a