A client came back with a last-minute change, and GPT-4 saved the day.
Last month, I wrote about how GPT-4 can finally work with Adobe After Effect’s powerful scripting system. I’ve only used it for experiments until last night. GPT-4 helped me create a custom script to meet a tight deadline.
The Dilemma
I was part of a team making a Mother’s Day video for a well-known consumer brand. We created a script and painstakingly cast a voiceover actor. The end product was nearly three minutes long.
Unfortunately, we received a last-minute change from the client at 6 pm yesterday. They wanted to kill the voiceover and replace it with animated text. This wouldn’t be a problem, but somebody was going to need to stay after hours to animate everything. Not ideal when we’re crazy busy and had an early call time to make matters worse.
The text couldn’t read like subtitles, but it also couldn’t be overly animated over such a long video. We needed something elegant and smart. We also needed to copy and paste each line from the script into text layers within After Effects, which was going to take a while.
The Solution
I asked OpenAI’s GPT-4 to create an After Effects script that would allow us to dump the entire three-minute voiceover script into a text box and have it do all the work for us. With some tinkering, we got it working in a matter of minutes.
The original video was sitting in a Premiere Pro timeline, so we just needed each line of text to last for a couple seconds. The timing would get re-done in Premiere Pro anyway to match the original voiceover.
GPT-4's script was able to chop up each line into individual text layers, and animate in the position and opacity in a pretty graceful manner. It worked with whatever font settings we had been using before.
In about 30 minutes, we had the entire project finished. I reckon we saved about two or three hours last nightm particularly because we could see how the animation would look in Premiere Pro within minutes and tweak it from GPT-4.
The Future
If you’re not already peppering in some AI over your motion graphics work, you may get left behind. Why not leave more time for creativity and stop putting redundant tasks in the way? In fact, there’s a plugin that brings ChatGPT into After Effects itself: Klutz GPT. It’ll make things faster if you’re creating ad hoc scripts and expressions. No more copy and pasting from your browser. Users can paste expressions onto multiple layers with ease.
What’s particularly helpful is that Klutz GPT will automatically feed an error back into GPT to fix bugs. In this article, hours were shortened to minutes. Now, minutes could be shortened to seconds.
However, this may not bode well for authoritative forums, whose questions and answers are where we used to get help with After Effects expressions and scripts. If users stop posting on Creative Cow, where will the next generation of source material come from? Food for thought, anyway. I can only hope these forums remain as spritely as ever.
So instead of spending 2-3 hours doing the work manually, you did it in 30 minutes via AI. That sounds great. With all that time saved, are you giving the client a 75-85% DISCOUNT for the time you didn't spend, or is the client being charged the same amount as if it was a manual edit? This is an important question because AI is going to create a whole new billing scenario in the industry. For example: a new version of "why should I pay you for an hour's work to do a mug shot when you did it five minutes" which has been an argument for decades.
AI is already making an impact in cost reductions, and not for the better. There was a story done a couple days ago about how voice professionals are seeing their workload being reduced because clients are using AI to create virtual voices. The hardest being hit are those who do audio books. Instead of hiring a person to spend hours and hours in the studio reading a book into a microphone, AI can create a verbal version in a matter of minutes not only in English but in different languages too. With AI one could have the distinctive voices of James Earl Jones or Dolly Parton do the speaking without hiring the celebrities themselves or celebrity impersonators.
Did you tell your client you used AI to meet their deadline and how much time it saved?
The client pays for a specific service delivered to a specific standard. How that service is delivered is irrelevant.
Perhaps the agreement was time based, perhaps it was outcome based. We don't know.
Your argument against AI is not without merit, but perhaps you don't recall that similar arguments were used against computers in the early 80s. Like then, what you're faced with is disruption that can be good or bad, depending on how the disruption is managed/applied.
What a truly terrible take this is. That's like asking an artist why they charge so much for their art if it could maybe take them only 30 minutes to draw. You're paying for the service and skilled individuals to provide it.
GPT made a last minute change a little easier, but it still took fine tuning and the skills of a professional. As well as the professional who licensed the Adobe tools
Not harsh, maybe salty, maybe misguided, maybe ignorant. But it'd have to deliver some kind of hard truth to sound "harsh" to anyone's ears.
I agree that this is misguided. "AI" tools are one of the best things to happen to freelancers and small business owners. So far "AI" has allowed me to take on more clients and turn around emails, proposals, copy and prep creative work much faster. Outsourcing and delegation are critical principles of any business at scale, likely the future of "AI" will be allowing teams to accomplish more in less time rather than fully replacing the team. Employees on the other hand ...
what a ridiculous take. that's like saying, "you used lightroom to do hundreds of pics instead of doing one at a time in photoshop. are you going to give them a 80% discount?" or you decided to share a folder of pics over the Internet instead of walking the photos over to the client and you're expected to give the client a discount for deciding not to walk to the client. AI is merely a tool to assist you. It can't totally replace you. We used to do math with things like an abacus or on paper. But now we have a calculator. Tools are there to help. They don't replace. And if they do replace, you have to evaluate what it is that you really do of value.
Times change. In the old days, we had milk men drive and deliver milk bottles from house to house. This was before the advent of refrigeration. When refrigerators took off, the milkman job was basically destroyed. But we move on. Now we look back and we think how crazy it is to employ someone to drive milk from house to house when you can just buy it yourself and keep it in the fridge for weeks.
I'm neither naive nor harsh, but am taking a realistic view of the business and ethical sides of AI to make a point. I absolutely agree that billing should be based on the project whether it takes five minutes or an hour -- it is the product that has worth; not the time it takes. But then, being able to produce quickly can warrant a higher rate versus a novice who might need several hours. This is how/why attorneys vary so much from $200 per hour to $1,400 and more.
I've been teaching college photography courses for over a decade. How to do things is not my prime focus; it is drilling into the students that photography is a "business" first IF you want to not just survive but be prosperous where you can afford to raise a family, send your kids to college, take vacations, save money for retirement, etc. Yes, AI is a new tool and way to do things, but it can also become a danger by changing perspectives on workflow and what to charge for it.
Here is a quiz I give to students on the first day: Their answers are in (). Companies like MasterCard, Visa, Citi, Discover, etc. offer credit cards to people. What are the goals of these groups: Ford, GM, Toyota, Honda, etc. (they make cars) NY Times, Washington Post, Chicago Tribune, etc. (they publish news stories) Nikon, Canon, Leica, Sony, etc. (they make cameras) And so on.
All those answers are flat out wrong. I didn't ask what they do, I asked what their goal is, which is the same as photographers. Their goals are exactly the same -- to make money. And that is done by learning how to do your work better than anyone else and efficient enough to be financially successful.
The problem that I see with AI is while it is a game changing tool and be an asset to photographers, it is also going to be used by others who will exploit the market no matter who gets hurt all to get the biggest share of the money involved.
On a similar note, look what has happened to licensing over the decades where it went from limited usage for X money, more usage for XX money and so on, versus today's contracts that give clients perpetual use in any form whatsoever via a royalty-free stock photo or a photographer who shoots a picture and doesn't realize what he/she is doing. Back in the '70s when I started, agencies represented photographers and fought for good licensing arrangements that benefited both of them. Many have now morphed into companies whose prime goal is to generate revenue for themselves with photographers getting pennies for royalty-free sales.
AI is going to change that where it will generate new imagery to the point that professional photographers are no longer needed in some markets. There's a story out today about actor Tom Hanks saying the way AI is going he will still be making new movies after he's dead.
There are different ways to look at this. Right now I prefer to see these generative AI services as just another tool in the toolbox. Not everyone will adopt this, or even be comfortable with this. But I am paying careful attention to what people opine about it.
From a human's perspective, generative AI is to photography (and many other professions, like IT) what robotics was to manufacturing, or even what steam engines were to manual labor. It is of that magnitude. It is a powerful tool that has the potential to provide its users with a huge jump in productivity. To me what the author described doing in this article is an example of productivity increase.
There are examples of AI-produced images and videos, threatening to eliminate the humans from the creative equation, but this article is not about that. This is an example of using a tool to increase productivity. It is important to distinguish between these two things. And yet of course we certainly have to acknowledge both of these things: tool for increased productivity and/or something that threatens to render humans obsolete in some regard, just like the locomotive engine and the robot. The reality is that with all of these things, humans are still part of the solution.
I see there being no more moral dilemma to using this tool to increase productivity than there is for using a computer instead of a typewriter or buying butter at the supermarket instead of buying a cow and a field and a barn, milking the cow and churning the butter.
BRILLIANT! I didn't know it could tie in to after effects! I'll definitely try this
What a great use of AI Stephen, thanks for sharing the idea.