Prefer to watch instead of read? Here’s the full breakdown on YouTube:
I run a video production company here in New York City, and I've been very hesitant to bring AI into my workflow. Then, on a recent paid client project, I ran into a problem where AI was really my only option. Thanks to Higgsfield AI, I fixed it in a couple of clicks, and I want to show you exactly how.
Full disclosure, Higgsfield is paying me to make this video, and they gave me credits to test their platform. They have no say in what I used those credits for, what this video is about, how I used the tool in my workflow, or what I say about it. I made that clear with them from the start. The Higgsfield link in this post is an affiliate link, so I may earn a commission if you sign up through it.
Why I Was Hesitant About AI
I do think there are people who use AI to steal other people's work, and I'm not a fan of that. But I also think that, used properly, AI can be a useful tool in your workflow. It should never replace you as a creative. It should help you create more. If you want my fuller thoughts, I wrote about where I stand on AI in video production.
If you want to see a higher-production example of what these tools can do, go look up what Sam Kolder did in his Higgsfield video.
The Problem: Amenities I Couldn't Film
The project was a real estate video for a building, and two of its amenities were off the table.
The first was a shared gym. If you don't live in New York, that's pretty common. Residents can use the gym, but it's also open to the public through memberships. Because it's open to the public, the building can't close it during business hours for a film crew, and if I filmed there anyway, I'd need a release from everyone who happened to be in the frame.
The second was the outdoor amenities. A nor'easter blew through about four days before our scheduled production date and damaged them. The repairs would take months, and the client didn't have months to turn this video around.
That left me with eight still photos from a photographer, which I didn't take myself. I'll be the first to say I wish they were better, and I wish I had the RAWs. But this is what I was handed, and we had to get something out for the client. And honestly, still photos sitting on a timeline are extremely boring.
Prepping the Photos Before Higgsfield
Before the photos went anywhere near Higgsfield, I did a basic color grade in Final Cut Pro so they'd match the rest of the footage. Without it, the stills came out oversaturated and way too contrasty, and they didn't fit the vibe of the video at all. It isn't perfect, but it got them as close as I could. Then I exported all eight as still images and put them in a folder named "for edit."
Side note: yes, I cut this project in Final Cut Pro. I've been on DaVinci Resolve for three years and I've been tempted to switch back, so you'll probably see a video about that soon.
How I Prompted Kling Inside Higgsfield
I'm not skilled enough to animate photos like this on my own, so this is where Higgsfield came in. I uploaded all eight images and set the clip to 10 seconds at the highest quality, since that's about how much footage I needed to cover that part of the timeline.
The biggest thing to remember is that you have to be very specific when you prompt these tools. Kling, the video model I used inside Higgsfield, really wants to create a story. That's great for some projects, but here I wanted each shot to stand on its own and still feel cohesive with the others, with no transitions that wouldn't match the rest of the video.
I also tend to ramble, so I described exactly what I wanted to my AI chatbot and had it turn that into a clear, concise prompt for Kling. The prompt asked for a multi-shot video of eight shots of equal length, with each reference image as its own shot and a hard cut between every shot. No transitions, morphs, or blending. It called for a slow push in on the first shot, a slow push left or right on the second, and so on. Every shot had to match the reference image exactly, with no color grading, filters, or style, and the biggest line was to add no people, characters, faces, logos, or text. I wanted it as bare bones as possible, like footage I shot myself, because I handle the effects in the edit. I just needed clean source clips.
Dropping It Into Final Cut Pro
I hit generate and waited, and it was exactly what I was looking for in one go. It did add music I never asked for, but I removed the audio, downloaded the video, and overwrote it onto the primary storyline right over the boring section.
From there I used Final Cut's detect edits feature to cut the clip at each shot, which let me move the individual shots around the timeline. To my knowledge, there wasn't really a way to do this kind of thing before tools like this existed, and it's an extremely useful way to work.
So How Did It Turn Out?
Does it look as good as if I'd gone and filmed it? No, and I think a lot of that comes down to the source photos I was given. But it's way better than what I had before, which was basically nothing. It took one prompt and one generation, with no testing, and it used 10 credits.
Final Verdict
I get why people are hesitant about AI, because I was one of them, and this was my first time ever using Higgsfield. After using it, I get why a tool like this exists and how powerful it is. If you use it as one tool in your creative tool belt, and not as the thing that replaces you, it can make real sense and feed your creative side instead of replacing it.
If you want to try Higgsfield yourself, click here to get started.
I want this to be an open dialogue, and I think it's one people need to have. These tools are here and they're powerful, and learning to use them the right way is going to make everyone a better creative. If you have questions, leave a comment on the video.
If you want to see my full breakdown, the full video is available HERE
