How I Made a Cinematic Speaker Reel for Under $100 in Three Days

Over the past two decades, I’ve paid thousands of dollars to professional production companies to make my speaker reels. Hired videographers. Worked with editors. Licensed music. Waited weeks for revisions. Waited longer for final cuts. That’s just how it was done, and the work was good. But patience and budget were always a factor.
Over the past three days I made a new speaker reel by myself.
Three minutes. Twenty-four scenes spanning forty-five years of my career. Original score. Custom cinematography. A through-line that traces the kid at his first TRS-80 in 1980 to the AI keynote stages of 2026.
It cost me about eighty dollars.
And it took just three days.
I want to tell you exactly how I did it, because if you’ve been wondering whether AI tools have gotten good enough to do real creative work, the answer is yes, and the difference between “good enough” and “indistinguishable from professional production” is happening faster than most people realize.
The old way
A genuine speaker reel produced by a professional video company in 2025 falls somewhere between five thousand and fifteen thousand dollars, depending on the production company, the scope, and whether new footage gets shot. The lower end gets you a competent edit of existing keynote footage with titles and licensed music. The higher end gets you a custom production with new shooting days, motion graphics, and a polished cinematic feel.
Turnaround for that work is typically six to eight weeks. You’re scheduling around a production crew, an editor, a colorist, a music licensing process, and multiple rounds of revisions.
This is the standard professional path, and there’s nothing wrong with it. Production companies do excellent work and the people running them are skilled professionals.
But the math has changed.
The new way
Here are the tools I used:
Kling for video generation, animating still images into cinematic motion. About $80 in credits across roughly 120 generations.
Nano Banana (Google’s image generation tool) for creating still frames that didn’t exist in my archive and my YouTube thumbnail. Free at my usage level.
Suno Pro for the original score. $8 for one month of access.
Filmora for editing, color grading, text overlays, and final assembly. I already owned the license.
Basically, I had to outlay less than $100 for this project. The rest was my time investment.
But here’s what I think matters more: the tools didn’t replace the creative work. They replaced the production work. The thinking, the story structure, the choice of which moments to include, the emotional pacing, and the script for the visual transitions all came from forty-five years of knowing what makes a good story. The AI handled the execution. I handled the direction.
How the workflow actually went
Day one was scene planning and reference photo gathering. I sat down with the question: if I had to tell my career in twenty-four images, what would they be? I wrote out a list:
A teenager at a TRS-80. A young man at a CRT monitor in 1995 launching one of the first thousand websites. The day Yahoo bought my games company in 1998. The stage at Internet marketing events. The bookstore the day my book hit the New York Times bestseller list. The set of the world’s first internet reality show. The backstage moment when iFart hit number one on the App Store. London’s O2 Arena. A montage of keynote stages from Microsoft to the United Nations. Radio City Music Hall. The moment I first sat down with ChatGPT and recognized what was coming. The current AI keynote stage. The closing walk through a corporate campus at golden hour.
Then I gathered every real photo I had from each era. Some I had to dig up from old hard drives. Others I had to recreate from memory. For a few, I generated entirely new reference images using Nano Banana, but always with my own face, my own posture, and my own glasses. Generation and assembly of the videos began that day.
Day two was more generation and assembly. For each scene, I gave Kling a starting frame and an ending frame, plus a prompt describing the motion between them. The prompt is the script. The frames are the bookends. The model fills in everything in between.
A typical prompt looked like this:
“A man in his early 60s walks confidently onto a brightly lit theater stage from the right side of the frame. He wears a vibrant blue and white patterned shirt and beige pants. He holds a wireless microphone. Behind him, a massive LED screen displays a static image. He walks with purposeful strides toward center stage, stops, and brings the microphone to his mouth. The camera follows him in one unbroken move.”
Six seconds of video, generated in about ninety seconds, costing roughly twelve cents in credits.
Most generations didn’t work the first time. AI image and video tools have specific failure modes you learn over hundreds of attempts. Hands come out wrong. Faces drift between frames. The model wants to animate things that should stay still. A person on a video screen behind you will start walking even though you wanted them frozen. Compound actions confuse the model. For example, getting into a car and closing the door in one shot is too much. Body proportions shift toward something that doesn’t quite look like you.
The fix for most of these problems isn’t to argue with the model. It’s to route around the problem. Two simpler shots instead of one complex shot. Static screen content composited in afterward rather than rendered in the original generation. Match cuts on body motion to hide impossible transitions.
Working with AI creative tools can look simple. But it’s not always easy. You see, the skill isn’t in the prompts. The skill is in editorial thinking. It’s knowing when to fight the model and when to find a different angle that the model can deliver the way you envision.
Day three was more video creation. But then there’s music, text overlays, color matching, and final assembly. I generated several variations of cinematic score in Suno, picked the strongest one, used Suno Pro’s stem separation feature to pull out individual instrument layers, and edited the score in Filmora to match my visual pacing. I wanted the drums to drop out for the ChatGPT discovery moment, strings swelling for the keynote crescendo, and a warm resolution for the closing walk.
Text overlays for each scene followed a strict template: year in the upper-left corner, milestone description at the bottom, consistent white text with thin black outline. Once I built the first overlay perfectly, the other fifteen took ten minutes each.
I started on Saturday. By Monday evening, the video was complete. Not a single actual frame of archive video was used. Every second was rendered from an original still image or an AI-generated derivative.
What this actually means
The headline reads like a humblebrag, but I think the implication is more interesting.
For most of the last thirty years, the gap between being able to imagine a piece of media and being able to produce it was enormous. You had an idea. You needed money, time, equipment, and skilled professionals to bring it to life. The divide was the entire production industry. You needed videographers, editors, composers, colorists, designers, the whole apparatus.
That has not completely disappeared. But it has shrunk dramatically for anyone willing to learn the new toolset.
I made a video this month that, by any honest comparison, looks like it cost between five and ten thousand dollars. The actual quality is high enough that I’m comfortable putting it on my speaker page, sending it to bureaus, and using it as the front-line asset in booking decisions worth multiples of that.
This doesn’t mean professional video production is dead. There are kinds of production AI cannot do well yet. Live event filming, complex multi-person scenes, and anything requiring real cinematography in real environments continues to secure work for the high end of the production market… for now.
But the middle has been compressed. The kind of project that used to require a small production team can now be done by one person with intent, taste, and three days.
If you’ve been waiting to make something, whether a speaker reel, a product video, a documentary about your work, or a film about your family, you don’t need to wait for budget anymore. You need a clear idea of what you want to make and the patience to learn the tools.
The cost barrier is gone.
The judgment barrier remains.
It’s not that AI is making creative work cheaper. The work isn’t cheaper. The execution is cheaper. The direction, knowing what’s worth making and how to make it land, that part has never been more valuable.
Forty-five years on the front lines of technology revolutions taught me that the same pattern repeats every time. New tools democratize access, the floor rises, and the differentiator shifts from who can afford to make something to who has the judgment to know what’s worth making.
We just entered that phase for video.
If you want to see what I made, three minutes that span forty-five years and six technology revolutions, here is the video.
The only question remaining is, what will I make next?



