Sora’s Videos Look Amazing — Until You Try To Use Them
A fun experiment that shows just how far AI still is from real-world practicality
I write about leadership, education, AI, and why teaching critical thinking about it is more important than ever.
Please hit the heart ❤️, restack 🔄, subscribe 📨, and all that jazz to help spread the word!
🙌
I finally got access to Sora video last week and decided to try it out and have a little fun with it. I started by throwing in a few quick ideas. I always start with tools like this by giving minimal scaffolding just to see what comes back. You can really tell a lot about certain biases based on doing that.
So, I threw in a quick prompt asking it to create an AI literacy commercial or video. I told it to base it on me as the person saying it, but it just created random people. What was fascinating, though, was that the people it showed were mostly people of color. I didn’t ask it to do that.
AI literacy is brand new, and we all need it regardless of race or gender. But my guess is it’s feeding off the word “literacy” and assuming who might need help with that. You can fill in your own blank, but I found it interesting.
After that, I gave it more specific input, but I still wasn’t getting what I wanted. Six or seven tries later, I went to ChatGPT, told it what I was looking for, and asked it to create some directives I could give Sora.
It took a few tries, but I finally landed on a couple I really liked. I took one, dropped it into Filmora, did some edits, and created a final version that I think looks pretty darn cool.
Now, what are my thoughts on it?
It’s fun to play with. I even created a video of a roller coaster, which was a blast. These could be awesome little videos for professional use because they are quick, creative, and easy to make.
Now the key word in that sentence is “could.”
There are ways to remove the watermark. I actually went into my Filmora and played around with doing just that. I did leave it at the beginning where the moon is and you can see it flickering there.
I could have removed that too, I just chose not to. It’s not as easy as it seems because OpenAI puts the watermark all over the place, popping up throughout the video. You have to go in and work through many different areas to remove it.
What I also noticed is that it caused flashes of discoloration in the video. So then I had to cut snippets, shorten, and elongate. When all was said and done, it took me well over an hour just to remove it in a 10-second video. Keep in mind, that was after the time it took to do the initial edits I had already done.
Maybe there are easier ways. I don’t know. But in my Filmora, it was not that simple, and I had to do a whole lot more editing to make it look right.
What I learned with the edits
Well, unless you really want to spend significant time fixing the videos or you have a tool that’s way more advanced than mine, I’m not sure how worth it these are going to be to produce if you need the watermarks removed.
Plus, it gets really sticky removing those. On something like this, there’s obviously no harm in it. But when people start removing those watermarks on videos that are deceitful or cross a line, that’s where the problems begin.
There have already been calls from people like Robin Williams’ daughter, pleading with people to stop bringing her father back to life through AI. Martin Luther King Jr.’s daughter recently spoke out about it too.
That’s the part of this technology that’s really challenging to justify, no matter how impressive the tool looks on the surface. When people use it for these types of things, it has serious consequences.