The AI Travels in Venice exercise serves a number of purposes not the least of which is learning how to predictably and reliably master the suite of tools I used to create the piece.
The basic premise is simple: using only photographs and AI tools, make a short travel video like any number of other travel videos I made before with analog tools then digital ones. Now I'm using AI tools to fly virtual drones through 2D space that's transformed into 3D space.
Probably the first thing I learned during the process is how easy it is to trigger a time lapse shot and how convincing is such a shot even at the hands of AI.
The second thing I learned is a tiny step toward pushing me to prompt for more than camera moves. For example, twelve seconds in I asked for a slight zoom with a pan toward the nearest mask on the wall. What I got instead was that nearest mask on the wall turning toward the camera as if someone wearing the mask was turning toward the camera. It's a subtle, cool, and slightly odd effect that I didn't ask for...
But it's better than what I asked for and joins the ranks of prompts I should've written.
The lesson was enough so that I've begun to ask for what I need and/or what I want. For example, the shot at thirty-six seconds covers a bit of narration about romance. With that in mind, it seemed the shot would be better, would serve the narration perfectly had the shot been taken during golden hour. So I asked the AI to apply golden hour lighting to the shot and the results is incredible especially on the waves to the left of the couple.
Again at a minute twelve in there's a couple at the far end of the alleyway who I thought would come across more authentically Italian if at least one of them was speaking with their hands. So I added that to the prompt and BAM. The one facing the wall from the opposite side of the table was speaking with their hands.
Also, in one iteration of this shot, the sun flared into being once the camera move cleared the first arch. I liked the effect so much that I asked for it in every subsequent prompt so that I wound up with both the hand talker and the sun flaring into being.
One thirty-nine is the AI calling the shot and me taking advantage of it. The original photograph is the silver orb with reflections of people in it. So my prompt was simply to have the AI camera push in slowly to the reflections.
And yet the AI camera wanted to see what was up and to the left, so... up and to the left the AI camera tilted and panned to reveal a woman in costume enjoying this moment as well.
Better though. And if not better.... intriguing.
At two minutes twenty seconds we have a gondolier with his gondola next to a building along the canal. The gondolier seems to be in waiting mode with the handle of the gondola's long wooden paddle in one hand.
By this point in the script, our host is talking about being serenaded so I find a bit of music, a song in Italian where, at one point, the singer sustains a long note. This way I can fake lip sync without having to account for individual words. So I prompt RunwayML accordingly and not only does the gondolier hold a note for me... before he does, he whips out a large black microphone from inside the handle of the gondola's paddle.
It takes a few tries to get the motion and timing right for this one moment of music... not too many, though, before I get one that's perfect. A true gift. ❤️❤️❤️
By two minutes twenty-six seconds, I'm all in. You see, at this point I have another gondolier who's waiting, standing in his gondola. Back in the day, when I was on shoots and my friend Tom saw people like this he'd catch their attention, smile, and wave. Inevitably they'd smile and wave in return... and so a lot of the travel shows Tom photographed include a lot of happy locals and tourists. So, as an additional nod to my experience on the road with Tom, I wanted this gondolier, this Italian man captured in a photograph, to smile and wave.
"Gondolier turns to the camera, smiles, and waves."
And darned if he doesn't do just that. No fancy camera work. Just a guy in a photo who turns, smiles, and waves. It's all a touch hesitant but still. It impressed the hell outta me. Seriously.
So that's a thing for me now. Regardless of what I believe the AI can or can't do...
Straight up. That actually takes a lot of getting used to because I operated most of my professional life within the boundaries of hardware and software. My creativity was restrained by what I knew for a fact was possible or not possible.
Those boundaries are, at the very least...
So there it is: Ask for what I need...
In great detail if the spirit moves me.
Or if I just feel like it.
The last most obvious thing I learned from this experience is that you can't ever ever ever, you can never go wrong with sunset shots.
They're just damned magic.
Photographs Unsplash & Pixabay GenAI Adobe Video RunwayML Script DeepSeek Narration ElevenLabs Music Pixabay SFX Pixaby
