How to Use AI to Build Your Audience for Your Documentary, Feature, or Short Film
by Carole Dean
I had an enlightening interview with Patrick Martin on the Art of Film Funding Podcast. Patrick is an AI expert, and the Co-Founder and Head of Technology and Development for AIFocus.net.
Patrick has been helping us at From the Heart Productions to stay ahead of the curve with AI technology, and to understand how these innovations can help filmmakers.
In the first half of the interview, Patrick walked us through a system to use AI to help optimize YouTube videos to reach more traffic. YouTube is an important platform for filmmakers to build their audience by sharing films, clips, trailers, interviews, and a wide range of other supplemental.
Patrick’s process uses three different technologies, Otter.AI, ChatGPT, and vidIQ, to determine the best keywords and to write descriptions for videos that will achieve high visibility on YouTube. This same process can also inform us of the best topics to focus on for future videos based on what people are searching for. There is another program he discusses called DALL-E which can be used to improve your video content. I’ve taken the highlights from that interview to prepare this guide.
Otter.ai: Accurate Transcripts for Video Content
The first step is to get an accurate transcript of your video or audio. Patrick says “there’s a system called Otter.ai that is 99.9% accurate, especially if you train it, meaning that you give it a lot of the same speaking voice, that will provide you with a complete transcript just by uploading the video. So now you’ve got a transcript that’s extremely accurate.” The Pro Plan is about $17 a month for everything you need.
ChatGPT 4 and 3.5: Choosing the Right AI Tool
“Now you can take that transcript and off you go to ChatGPT. We recommend that you get yourself a paid account. It’s only $20 a month, but that gives you access to both ChatGPT 4 and 3.5. What’s the difference between the two? 3.5 is a bit like a regular horse, and ChatGPT 4 is like a racehorse. It goes faster, and it wins more races. But if you don’t know how to ride a horse, don’t jump on a racehorse first.”
Multi-Prompting: Boosting Efficiency with Context-Specific Questions
Regardless of which version of ChatGPT you use, Patrick says you’ll be faced with what’s called a ‘prompt engine.’ The prompt engine is a very important thing because it’s not just going to give you a simple sequence of answers like Google. It’s going to do work, real work.
“What we want to do,” Patrick pointed out, “is to put that transcript from Otter.AI and cut and paste that right into ChatGPT. Then ask it to provide a set of long-tailed keywords.
“A long-tail keyword is a combination of keywords. It’s a fancy way of saying, well, here’s a sentence with two or three keywords in it. 80 to 90% of search is governed by these long-tails, meaning there’s lots more variations of combinations of keywords than there are single keywords. The great majority of traffic comes from long-tail keywords.
“ChatGPT will spew out 15 short sentences of long-tail keywords. What we’re really doing is called multi-prompting. That means start by doing one thing and obtain a result. We’ve taken an audio tape, and now we have text of a transcript. We then take that transcript from the AI and off we go to GPT4, which is much more of a large language model.
“Now what we’re doing is we’re going to get another artificial intelligence, and this is one that anybody that’s on YouTube should be using, because they’re pretty much a monopolist on this. They’ve got all the data, and it is called vidIQ.”
vidIQ: Leveraging AI to Identify Low-Competition Keywords
“vidIQ is a different type of AI. It’s an AI that is looking at all the data that is flying around in YouTube, and it’s saying, well, we’ve got access to all this data, and we can tell what keywords are highly competitive.”
We talked about some search terms that might be appropriate for From the Heart to use for our videos as examples. But the same principles will apply to whatever your film is about.
“For example,” he explained, “‘movie making’ is a highly competitive keyword that has 500,000 people looking for it. It’s extremely high competition. So, the likelihood of ever getting a top position for anything you post is virtually zero, and you’ll be on page 96, with no traffic.
“If on the other hand, you ask for terms that are less competitive for argument’s sake, film funding for documentaries, you happen to be making a documentary. So therefore, you are looking for ‘film funding for a documentary.’ Well, we see there’s not that many people looking, maybe 2000 or 3000.
“The amount of competition, meaning the number of people that are competing with you for that term, is low. Well, that’s important to know, isn’t it? So now you can take the results that ChatGPT 4 has given you, all the long-tail keywords that are associated with the transcript, and you can put them into vidIQ and test which of those keywords are effectively low competition with a reasonable amount of traffic.”
What we’re trying to do here is find the “low-hanging fruit,” which are the “keywords that I can actually get a top three, top five position and really get some traffic for people that are interested in what I am doing.”
Crafting Powerful Titles: Combining Human Intelligence with AI Insights
Now, you can use your human intelligence to create a title for that work. “These types of technologies use traditional skills. You must be able to construct a meaningful, important title for your video once you’ve got the information on what is likely to be the low-lying fruit. Using your existing skills as a writer, screenwriter, editor, or as a designer, all of these tools are now available to speed up the work and to produce a quality of work that’s much higher with much less time.”
I’ve also found that ChatGPT is a great tool for coming up with titles as well. I put my transcript in the prompt, and asked ChatGPT to create a title and to optimize for the long-tail keywords I wanted to focus on. It came up with some terrific options.
Writing Video Descriptions and Chapter Segments with ChatGPT
The next step is to put our title and transcript back into ChatGPT and ask it to come up with a good video description for YouTube. It will come up with some great keyword-rich text that will help your visibility. The other thing you can ask ChatGPT to do is to divide the video transcript into chapters or sections if your video is structured that way. It can help you come up with titles for these chapters as well.
“What we’re doing is called ‘prompt engineering,’ which asking very specific questions in a series to get very important information.”
DALL–E: Creating Compelling Images and Avatars for Videos
Another thing you can do with AI to improve your videos is to use DALL-E, which allows you to create images from prompts. So, you can say “’give me a hobbit living in an oak tree,’ and it’ll produce a hobbit in an oak tree for you. There are other operations with AI that will then take that hobbit in an oak tree and create mouth movements for speaking for 10 or 15 seconds, realistically just sort of Disney standards of talking with what I would call a ‘locked off shot.’ No moving cameras, just a locked off shot, somebody facing camera or a hobbit facing camera.”
“Then you put in the text, and then the Hobbit will say, ‘I’m a hobbit that lives in New Zealand.’ And those are the technologies that are also extremely useful in certain important categories, such as news releases and all the various things where avatars are now being used by the BBC, Microsoft, and everybody else.”
Democratization of Data: Harvesting Your Own Insights from Google
“Google became the biggest advertising agency in the world. Most people think of Google as a search engine. No, it’s, it’s an ad agency cleverly disguised as a search engine. That’s what it is. Everybody’s interested in Google, but Google’s only interested in itself because the more times you look at Google, the more money they make.
“And so, from that point of view, yes, they give you lots of free stuff, and AI has been used on everyone for the last 10, 15 years. They have been using AI to gather as much data on human beings that are using Google system as is humanly possible so they can sell it to advertisers. So, you know, if everybody’s worried about the government and having information on you, well, I’m not worried about that at all, but Google is another matter.
“The democratization of data has begun, and now the big boys are being forced to give up their data. So, to a certain extent, the work that we do is allowing businesses like yourself and other businesses that we work with to harvest their own data from Google, which Google allows you to get.” There are many types of data that can be gathered from Google. Patrick gave the example of “how many people found you by searching a map.
“A lot of people look at the map, they don’t look at search. And so with all these variations of how people found you, what did they do? Did they come to you with a train? Did they come to you with a plane? Did they come to you by driving? And how did they get there? And what did they do when they arrived? And so on and so forth. All that data’s available that every business, every school, every university can collect that data from Google. Very few do, but those that do tend to excel.”
Navigating the YouTube Landscape: Building Consistency and Subscriber Base
“Having your own YouTube channel and putting your flms and marketing yourself is the way of the future. If people are interested in your specific subject, whatever that may be, then they will subscribe. And the key is consistency.”
“90% of the traffic that you’re going to get is not going to come from search. It’s going to come from suggested videos.What is a suggested video? Well, first of all, let’s look at what most people do when they make a video. They don’t do anything other than upload it. And then they say one sentence or two sentences about it, and there it remains collecting digital dust until it disappears, because nobody’s watching. It gets 50 views or 60 views.”
“The reason that it’s important is not for you or for the person watching it. No. It’s important for YouTube to keep you looking at the ads. That’s why we’re here, right? To watch the ads. We’re not here to do anything else other than watch the ads. Google is not concerned with the content. It’s only concerned with whether the content is interesting enough to keep you watching on YouTube.”
“So, you’ve got a bunch of keywords such as ‘documentary filmmakers and funding’. And you then look at the top of that particular keyword, ‘documentary film funding.’ And what do we see here? We see lots of old videos, two years old, three years old, five years old. What does that tell us? It tells us there’s no competition, and we’re more than likely to get a top three position because of a very important term.
“That’s another lesson to be learned here. Recency. It means “when was this published?” Because the only thing Google is really interested in, if you really want to succeed on them is “am I a consistent publisher?” Being a consistent publisher means putting out a couple of videos at least a month, and some people are putting out two and three a day. They don’t have to be long. It can be 10 minutes, 15 minutes.”
“The point is that consistency builds up what we call channel authority. And what does channel authority confer? It confers reaching for the higher fruit. So at the beginning, you can only go for keywords that are relatively low competition, and maybe not so many people looking for them. After you’ve got your channel authority in whichever niche you happen to be in, then you can become visible with those more competitive keywords.”
Conclusion: Embracing AI to Optimize Your YouTube Success
AI is here to stay, and we must embrace these technologies to stay ahead of the curve. There are wonderful tools here to help us to get our film and our voices to a much bigger and more engaged audience, and I encourage all of our filmmakers to take the time to learn more.
Listen to the entire podcast in the link below:
Carole Dean is president and founder of From the Heart Productions; a 501(c)3 non-
profit that offers the Roy W. Dean Film Grants and fiscal sponsorship for independent filmmakers.
She is creator and instructor of Learn Producing: The Ultimate Course for Indie Film Production. 26 classes which will teach indie filmmakers how to produce their films.
She hosts the weekly podcast, The Art of Film Funding, interviewing those involved in all aspects of indie film production. She is also the author of The Art of Film Funding, 2nd Edition: Alternative Financing Concepts. See IMDB for producing credits
