In Two Years Every Gym Will Have the Same Images
How I use AI on real jobs: thumbnails from unused footage, a video editor that runs the multicam while I do something else, and the part I still have not cracked.
Business owners are all making their flyers in ChatGPT now. Same brushes. Same icons. Same fonts. Put four of them next to each other and you cannot tell one business from the next.
I understand why it happens. The owner is busy. There are a hundred things to do and something has to go out today. So it is create a flyer for this event, go, make no mistakes.
In two years, when every gym has the same images, it is going to be very difficult to differentiate.
Somebody looking for a gym pulls up you and the one down the road and cannot find the difference, because your visuals are literally everything ChatGPT told you to do. That is the part nobody is thinking about while they are looking for the easy solution.
16:50
Before you post the next AI flyer, put it beside three from your competitors. If you cannot tell whose is whose, it is not doing anything for you.
How I actually use the image generator
My last podcast with Aaron ran 55 minutes. For the thumbnail I did not shoot anything new. I went into Premiere, saved a frame out of the footage, and took a few options. The screenshot is not as clear as a photo, so I took it into ChatGPT and told it what to clean up. Remove the cables. Remove that part. Leave the eraser, the whiteboard, the markers, the TV.
One thing you have to say every time is do not change the faces. If you do not say it, it will change them. And it can only work with the detail that is in the frame, so when the original was soft it missed things like a ponytail.
3:10
Then the text goes on in Canva, not in ChatGPT. I can add text in ChatGPT now, but it will not place it where I want it. I wrote the line myself. I told it the font, the indigo color grading, the color value, the aspect ratio. That is the whole thing with these tools.
The advantage is you telling it what to do.
Same trick on a client job. I needed a thumbnail for a podcast and I had a good photo of my client but nothing of her guest. The best image I could find on Instagram was way too small and already had a background, so it would have looked pixelated in the thumbnail. I screenshotted it, told ChatGPT the resolution and the aspect ratio I needed and to remove the background, and it came back gorgeous. Then I color graded it and dropped it in.
Treat the generator as a repair shop for assets you already own, not a replacement for your own idea. You bring the frame, the text and the colour value. It does the cleanup.
Save a frame out of footage you already shot instead of booking a new photo.
Tell ChatGPT exactly what to remove and what to leave, and add do not change the faces.
Write the headline yourself and set the text in Canva so it lands where you want it.
Add a line to your prompt telling it not to always agree with you.
The editor runs while I do something else
As soon as we finish recording, the footage goes into a folder on my computer and the agent takes over the multicam. I run it through Claude Code because I like it, but there is Codex, there is ChatGPT Atlas, there is Antigravity from Google.
5:45
When the edit is done it builds what I call a package. It comes back with a recommended title, a set of other title options, and the text for the YouTube thumbnail, pulled from what we actually talked about. It even flags things to double check before I use them, like numbers.
I asked it to time itself on one episode so I could show people. It started at 6:43. The reel was ready at 7:24. The artifact came back at 98 minutes total, about an hour and 38 on the podcast. Some of those numbers I do not fully trust, it told me two hours once when it was not two hours, and I think it is counting the gaps between my answers.
That edit takes me about two hours, and I am fast. The number on the screen is not really the point.
Those are two hours of my time. AI took over, so I can use them for something else.
Photo editing without opening Lightroom
I shoot Fuji and I love it, but I work with Canon, Nikon, Sony, all of them are great and they have different purposes. On a recent brand shoot the originals were already good. I asked the photo editor agent to show me different film grades, Portra, gold summer, the popular ones.
Now I can put those in front of the client. You tell me you like gold summer and I send you every image in gold summer without editing them one by one. For my taste the oranges in that one are too much, so I desaturate a little and stay with Portra. I can change the contrast, the sharpness, the dimensions. I was already fast at this. This is just another way to play with looks.
Give the client a choice of finished looks before you commit to editing. You are selling a decision, not hours in Lightroom.
It will always agree with you
Ask it whether you should put your money into something and it tells you that is a great idea. Ask it about doing something else with the same money and it tells you actually you are right. So put it in the prompt. Do not always agree with me. I have it in my training so it does that for me.
21:46
What I still have not figured out
AI image generation, and I hate that. Three years ago I predicted we would not need models anymore, or one model at most. The brand sends the outfits, I tell the agent to create the full photo shoot, give it poses, describe a blonde who is six one, describe the shapes. I create the images and it messes up. I have been in communities, I have asked people who know more than me for advice, and I still have not cracked the code.
What I do see is the photographer becoming the director. You describe the lighting, the lens, the camera, the scenery. Real models and real people still have value. Getting the exact pose and the exact lighting on a shoot is time consuming and energy consuming, so for putting concepts in front of a brand this saves time.
Six months of boring
None of this gave me a solution right away. The video editor took me about six months of going back and forth before it did its job. It was frustrating. There were nights I told it this is first grade type of stuff and you still cannot get it.
34:43
I am 46. I have made plenty of mistakes and the value I found is in staying in the boring thing for as long as you can. Most of us have the shiny thing syndrome, the next thing and the next thing. I do not even like the word consistency. You just have to admit it is boring.
Some days I have to numb myself. There is still so much work to do and if I stop to think about how unfinishable it is, I would quit. So I take the first thing off the list and I start. I do not let myself feel how small I am. This is not motivational. It is not about demons and carrying logs.
The people who get through it are not the ones who found a better mindset. They are the ones who keep looking down and go to the next task.
Why DRU exists
Content creators buy the cameras and the lenses and skip the business side. DRU is Daniel Ro University. Think of a technical degree that takes two years and gets you earning without spending a fortune on college, except you do not need money to start.
You get access to my agents, my process and my experience getting clients. Once three clients are paying you, I take a percentage. After 18 months you leave with five to seven clients and they are yours. The system can take you to a million in three to five years if everything is perfect, and not everything is perfect. You will find challenges and you have to be dedicated.
01
Direct the tool. You bring the frame, the headline and the colour value, and you tell it not to change the faces.
02
Buy back time, not output. The edit running without you is two hours you spend somewhere else.
03
Stay in the boring thing. Six months of back and forth is what a working system costs.
Stuck on a use case
Tell me what you are trying to get done in your business and I will tell you if AI is worth pointing at it or if you are better off shooting it.
Send me a text