Transcription
So the slow part of AI filmmaking is not generating the video all the time. It's everything that happens before it. Casting, character development, scene setup, shot planning, consistency, all that. That's where most of the time actually goes. And honestly, that's where a lot of projects fall apart.
But I think I found something that changes that because when you combine Claude Co-work with Higsfield, it starts to feel like a superpower. And if you're new here, I'm Kalo from AI for Real Life, and I've been deep in the AI filmmaking space, testing tools, breaking workflows, and figuring out what actually holds up in real projects, not just demos.
And right now, I'm working on a project in the background, a chubby checker trailer that I've been slowly building out whenever I have time. But that's the problem. I don't have a lot of time between client work, building out my course, taking my son to soccer practices, and everything else. I need a way to keep making progress without sitting there and doing all the tedious setup work myself.
So, I wanted to test something. I went into Claude Co-work and connected it to Higsfield and basically said, "All right, let's see how far we can actually take this." So, let me walk you through exactly what happened.
So, the first thing I did was I created a dedicated project inside of Claude Co-work for this trailer. And I already had a bunch of documents like beat sheets, script ideas, notes, all sitting in my Google Drive. So, I told it go into my Google Drive and pull all the relevant documents for this project and organize everything into a clean project structure. So, we've got a working database. And it did that and it was actually pretty fast. So, I was already impressed right there. Everything was organized and structured and ready to go. So, now instead of digging through folders trying to piece things together, I had a clean foundation to work from. All right.
So the next step I said based on the beat sheet, the script, everything that we have here, go ahead and build me out a detailed shot list. And it went through and did that shot by shot, structured, clean, ready to use. So now we're already skipping a huge chunk of manual work.
But then I took it a step further. I said, "All right, now take that shot list and build out detailed prompts for every single shot. And not just generic prompts. I want them structured specifically for cling style generations." So now we're not just planning shots, we're preparing them for execution. And it knocked that out too. Not a big surprise. So at this point, everything is still pretty cool, right?
But this is where it actually started to change things for me. I told her, "All right, now go into my Higsfield account and start building the actual visuals." Now, I had already created the main characters for this project, so I just went ahead and gave it the process. I said, "Use Nano Banana Pro because I have unlimited generations with my Hicksfield plan. And make sure you tag the character properly whenever needed in these prompts. Uh, use a 21x9 aspect ratio. Stick to these parameters and also only use unlimited generations and go through the entire shot list and start generating all the stills."
And honestly, I didn't really expect much from this. I left, went did some other things and I came back later and it had built out all the shots. And now all of them weren't perfect, but I was already 10 steps ahead of where I would have been. And that right there is where this whole thing started to feel different because it felt like I had the smartest intern in the whole world just knocking all of the tedious work out for me. So stuff that would have normally taken me hours was already there waiting for me to just review it. So now instead of spending all that time setting things up, I'm just coming in and selecting what actually works. And then I just do some refining from there. So that right there was just blew my mind.
So the next step after this whole process was adding in motion. So then I had it go back to Cinema Studio 2.5 inside of Higsfield and asked it to take all of those clips and asked it to take all those prompts that it had made for me in my shot list and start generating video clips from each shot. So to be honest, I had some credits in there to burn. So I was fine just testing this out pretty aggressively just because of the upside of it on the back end if this actually worked out. But what it did was it actually went through that shot list. It used the prompts and started building out the video generations for me.
So now I'm at the point where I've got the beat sheet, the shot list, the prompts, the stills, and now the video clips all generated. So my whole role here just shifted. Now I'm going through each clip and seeing what works for the film or for the trailer. And if something doesn't work, I just tell it what's wrong and we run it again. So instead of me spending hours setting up shots, characters, and scenes, that part is already done while I'm working on other things. And this part right here is the part that actually changes the game because now I'm not stuck in setup mode. I'm in selection and direction.
And I've done the same type of process on Freepick, uh, using 11 Labs, different platforms. So this isn't just about Hfield right here. This is about the tool that I use for this specific test. The bigger idea is this. You can offload an entire testing phase. So if you have unlimited generations with whatever plan you have, you can literally just let it run through variations, try different setups, build out options, and then you come back and just pick what actually works. And that's a completely different workflow. And honestly, this is just scratching the surface. This opened up my mind to what's actually possible here and how much of the tedious process I can actually remove. And I'm probably going to break this down in a much deeper way inside of my school community and show you exactly how I set this up step by step. But even though I thought about not putting this out here because it gives that sort of unfair advantage, I went ahead and put it out here because this is already changing how I work behind the scenes.
So for me, that's the real shift here because now the question isn't can I make an AI video. The question is how much of the process can you offload before you even touch your generations? Because if you can do that, you can move a lot faster. So here's the way I look at this. Right now, a lot of people are spending something like 70% of their time in setup and maybe 30% of their time actually generating. What this does is flip that and that changes how much you can actually produce.
Now, this whole thing is still early. It's not perfect, but this is one of those moments where you can see where this whole thing is headed. And honestly, this changed how I'm thinking about these tools moving forward because this isn't just another feature. This is like having a second brain or the smartest intern in the world helping you build.
So, I don't know. This has been my process. I just really wanted to share it with you. Uh, but I don't know, what part of your workflow would you actually hand off first? And how do you feel about this whole thing? Drop it in the comments. And also, if you made it this far, go ahead and hit like and subscribe so that you can catch up on all the new workflows and everything that's coming out around AI filmmaking. And I'm going to get back to building out more modules, guides, and breakdowns inside of my school community. So, if that's something that you're interested in, check the links in the description, and you can go ahead and get in early. and I'll catch you on the next.