Rich Tella Filmography



How I Plan to use AI in my Filmmaking

There’s a hell of a lot of attention on AI filmmaking these days, and, of course, due to the visual nature of film, almost all conversation is entirely placed on the generation of AI video. Regardless of how you feel about it, it is here to stay and will likely only get better with time.

While that’s fine for those without the means to create live-action, on-location films, for me, I’ve been slowly watching and waiting for a worthy assistant to show its head, so that I might indeed not have to deal with all the laborious tasks that come with independent filmmaking.

I recently came across this video about Notebook LM, and this made my ears, eyes, and mouth pop wide open. Finally, I saw the perfect way to introduce AI into my own personal style of filmmaking.

While at the time of writing, that tool is only available in the States, here’s how I intend to use it or something like it to enhance my workflow and even introduce potentially new ways to quickly identify components of my stories I would otherwise have likely missed or just not had the time to discover.

If you work with video, you will know just how much raw material there is to manage—from scripts to graphics, to music and voice recordings. If you’re like me and a tad bit messy more often than not, once a project is finished, it gets filed away on an external drive. If not labeled and cataloged properly, it will remain a nightmare to search for anything of use that may be required in future or current project.

Assuming, like me, you learned the hard way and on your most recent projects took the extra time to label and catalog each asset properly, congratulations! Now you have a library that is easy to search for specific content.

But what about making sense of it all?

What if you forgot something was perfectly relevant to something else in your latest project but cant figure out how it fits within a newly found context?

This is what I primarily want from an AI filmmaking tool: that additional brainpower so that I don’t miss any opportunities for enriching the story I’m currently trying to develop.

And so, given the opportunity, here are some ways I would like to make use of this new additional brainpower afforded by LLMs or AI or whatever you prefer to call this new superpower:

Querying all my previous production assets for relevant info applicable to my current production

While this may not work for everyone, in my particular case, the character drives the stories forward from one production into the next and often takes the form of a mini-series. Being able to instantly query all my previous production assets for relevant information applicable to a current production, even a specific given moment, would help ensure I don’t miss any opportunity to make the current scene, moment, or sequence even better.

This is something that has always hung over me, knowing that there was something that happened previously that is applicable right now in this moment, but I cannot for the life of me recall what it is exactly. With such a tool as mentioned above, I should be able to have a conversation with the entire back catalog in an effort to realize that which I have forgotten, and at the same time, find exactly where the heck it is and what type of content it is.

A tool that helps facilitate this in almost real-time would serve a real tangible benefit to my process. Given that if I were to try to do this these days, even though my libraries are well-kept, it would still take a lot of time just to try to identify what the hunch I have actually is in relation to.

Having a broad discussion with all my interview transcripts

With documentaries, you often end up with not only hundreds of hours of video content but also many hundreds, if not thousands, of pages of transcripts. Wading through these transcripts to find what you need can take a lot of time—time that you might prefer to use in other ways. So, you know what you want, but you still have to get in there and dig it out. That’s fine, but it’s also the stitching of those segments together in relation to what other people said in other clips or interviews that may or may not be applicable.

My hope is that I will be able to simply have a discussion with all the transcripts at the same time, querying for a specific perspective or moment that might be shared among many interview participants. I seek out patterns and applicable context, which may well be overlooked when doing this kind of digging manually.

Having a discussion with all assets regardless their state

Interviews are generally well-structured, often just a back-and-forth between the interviewer and interviewee, so the transcript isn’t so complex to query. However, if you were able to feed not just scripts and raw footage but also post-produced footage, which context can change as you move through the production process, this would allow you to have a discussion with all the assets no matter what state they are in.

In documentary filmmaking, being able to quickly cross-reference applicable story components, regardless of their current state in the process, would be immensely valuable for identifying mistakes. In journalistic content, it could also help facilitate fact-checking.

So while all the hype online continues to be around the visual aspects of AI filmmaking and video, I’m confident there will always be many people out there who still get a kick out of being out in the wild, shooting live-action stuff with their friends, especially in the documentary genre.

Personally I will be using AI-generated-video tools primarily for economic reasons—generating aerial shots, generic B-roll, etc. Given the cost saving effects It just doesn’t make sense to do it any other way.