Let's run an experiment. Can AI make a better video than a human? Artificial intelligence. Massively powerful computers. Chatbot. Human intention. Generating or creating content.

 

Everyone is talking about AI these days. It's revolutionary. It's exciting. It's also a little bit scary. It's going to affect every industry in time. But in the video production industry, are we at a place yet where we feel that AI tools can make you a video end-to-end? There's plenty of exciting tools out there, but let's really dig in and see how effective they can be to get a video in hand with little to no input from a human.

 

Let's start by feeding a really simple brief to ChatGPT. It's probably the AI tool that people are most excited about. I'm going to feed it a really simple brief. Why is making a commitment to zero carbon so important for businesses today? Ooh, it's just full on writing a script for me. Or it's rather answering the question. I didn't actually put in, write me a video script. I just gave it the brief. So it's kind of generating the material that it thinks I need to know--Environmental concerns, regulatory pressures, consumer demand, cost savings. And it's done, like within a matter of seconds. The instant gratification is crazy great. It's certainly quicker than giving a brief to a human and giving them some time to do the research and come up with what they need to feed into it. But let me see if I can get more specific. Let's get it to actually write a script.

 

It's writing, and ooh, this time it's even giving me its ideas for opening shots. Opening shot of a polluted city landscape with cars and factories emitting smoke. And then it's telling me what the voiceover should be, which is pretty great as well. Some of the language is a little bit clunky, saying thirdly, fourthly, probably, something a human would never do. The script isn't bad. It's not perfect. There's definitely things that I would change and it is quite literal in the way that it's interpreted my ask, but it's pretty detailed in terms of what it thinks I should put down for visuals and what I should be saying as the voiceover and who should be even saying what. So that's pretty impressive. Now let's move on. Let's see if We can find an AI tool to actually make this into a video for me.

 

I found this tool called Hugging Face. It seems that you just put in a prompt and it'll generate video for you. So that's cool, but I don't think you can put in the whole script. So let's just try a keyword. Clean air. Generate video. Okay, it's generating something. I'm in the queue. I am 15th in the queue though, so it's going to take some time for this to work. 

Ooh, what is this? Okay, so Hugging Face gave me something a little bit strange and not exactly what I wanted, but I think this model will probably learn as more users input different asks and it learns to output closer to what we are wanting.

 

Text to video is hard, but what about an image? I found on GitHub this tool called 3D Ken Burns and the preview video looks pretty cool. It looks like it can parallax any image, which is a great tool to have, but it looks like the way that you're supposed to use it is with a Python script. And I don't know the first thing about coding. I'm sure there's plenty of people out there who could totally use this. But for me, as an average person already, it's something I can't use, though it looks really, really cool. So let's find something else that I can actually easily use.

 

There's a tool called Steve AI that seems to let you just paste in a script and bingo, you have a video that you can actually customize. However, it's okay. Opening shot of a polluted cityscape with cars and factories emitting smoke. voiceover our planet is facing a climate crisis and businesses have a significant role to play in addressing this issue.

 

It's not great, but it's actually not terrible. It's pulling in some relevant clips that I might want to see, but again, those are very on the nose. It actually looks like Steve AI is pulling from Getty Images and is like reading the descriptions of those clips to see what is relevant to pull in, which is actually really clever and maybe will even save you some time if if you are trying to use this to get ahead and making a video come together more quickly. But if I were to feed this same brief and script to a human editor, they would be looking for a lot of other qualities that the AI tool just can't. They'd be looking for color space and tone and the movement of the footage and how these clips might sit together to create a visual metaphor. Let's continue our search.

 

So there's this tool called Murph AI, which lets you basically paste in your script, pick an AI voice, and there's a huge selection of AI voices to choose from. And it looks like it lets you go through their library of visuals to pick relevant images to match your script. Okay, so this is in the timeline. Opening shot of a polluted cityscape with cars and factories emitting smoke. Their UI is pretty easy to navigate. It should be pretty straightforward for anyone just stumbling upon this tool, but it kind of feels like a light version of editing in Premiere.

 

Not too long ago, Runway pushed the boundaries of generative AI with Gen1. RunwayML has had a lot of buzz around it recently as the latest and greatest AI tool in video. But can it actually make a full-on video for me?

 

Okay, next, let's try RunwayML. Everything you need to make anything you want. Ooh, AI magic tools. Yes, that's what I'm here for. We have Gen1. 
video to video. What on earth is this? I uploaded a really nice clip of this docu-style trailer that we did a while ago and it's created something that looks like a dystopian horrifying version of it. Interesting. I'm not sure if that was exactly how you're meant to use this but fascinating test. Train your own generator, image to image. Expand image, frame interpolation, erase and replace. These things feel like they can't really take my script and just spit me out a video.

 

There's a lot of interesting tools here that feel that they could be really useful, but they're geared more towards industry editors and animators. People who already have a working knowledge of how they want to edit and what they're trying to create, but just need a tool to quicken up a task for them. It doesn't feel like this is a tool that's for the average person trying to go from brief to delivery of a video.

 

So can AI tools actually make a better video than a human? I think right now the answer is no, or at least not yet. ChatGPT is by far the most powerful and interesting AI tool, but AI and video is actually still super desparate. There's lots of individual tools that can do one thing really well, but there's not quite one cohesive tool that can bring all of the different pieces together to just plop you out a video. There's definitely an opportunity for a stock site to feed their library to an AI model and have it become a more autonomous editing tool. Steve AI is the closest to this, but it isn't quite finessed. You still need to input to get it to swap clips and adjust the video to get it to a decent place. The speed though is remarkable. Literally pasting in the script to a final video in hand was 10 minutes flat.

 

So should you be worried as a human in this industry? The answer is maybe, but instead of being scared about how AI tools are going to be replacing humans, our stance today is let's try these tools. Let's not be scared of them and see how we can use them to free up our time and make our lives better.