Using AI in Video Production: The Writing, Not the Camera
Jan 30, 2023
AI's job in video production is the writing, not the filming. In this episode Chris Schwager works through what it had actually done in his own business: headlines, descriptions, social media replies, scripts, culling long copy, coming up with new topics. Not a frame of footage. What it saved was brain space, not production time.
Episode 137 of Video Made Simple, the show now called the Video Confidence Collective. Chris Schwager, Founder and CEO of Ridge Films, types into the tool live on camera and reads the answers out as they arrive, including the two that run out of capacity halfway through a sentence.
What can AI actually do for a video business?
It writes. Chris lists what he had been using it for and every item on that list is words, not pictures.
"So this is, AI for us has been about, you know, writing more succinct headlines, descriptions, social media posts, scripts, you know, abbreviating, culling, different things, coming up with new topics, expanding on ideas."
Two of those are worth pulling out, because they are the ones nobody plans for. Culling is taking something already written and making it shorter without losing the point, which is slow and thankless. Expanding on ideas is the reverse, turning a one-line thought into enough on the page to argue with. Neither is scriptwriting, and neither of them appears in the finished video.
Do you need to write a perfect prompt?
No. A short follow-up in the same conversation works, because the conversation carries everything you asked before it.
"It's actually learning what I'm, where I'm going and it's looking at my previous correspondence here and gluing this together for me"
What Chris typed next was closer to "what about for videos" than a brief, and he says on air that it is an incomplete question. The answer came back about scripts, captioning and video descriptions anyway. The usable version of that: ask the broad question once, then keep every follow-up short and in the same thread, rather than rewriting the whole brief each time.
How do you reply to a comment you do not want to answer?
Paste the comment in, ask it to simply reply to this, then pick one of the paragraphs it hands back and post that one. Chris did it with a run of odd TikTok comments he did not want to answer emotionally on behalf of Ridge Films.
He describes lying in bed wanting to go to sleep, and feeling obliged as a company to answer them anyway. So he stopped trying to write the perfect response.
"Pick a paragraph, put it back into social media and respond and be done with it. No energy, no brain space needed."
The value there is not that the reply is brilliant. It is that the reply you write in bed, mildly annoyed, on behalf of a business, is the one you regret in the morning. Handing it over takes out the part of you that wanted to argue.
Can AI generate the video itself?
Nothing in this episode makes a video. Every output is text: an answer, a headline, a description, a reply to a stranger. The camera, the room, the light and the person sitting in front of it were all still Chris's job, and that is the half that decides whether anyone believes what they are watching.
That split has aged better than any tool named in the episode. Equipment and software are the easy pillar. The message, the delivery on camera, and actually shipping something every week are the three that decide the result, and none of them can be handed to a model.
If the AI presenter question is the one you came here for, it has its own page: can AI video generators replace you on camera. For the scriptwriting side, read AI video scriptwriting: structure, speed and control.
Key moments
- 0:28 What AI was actually being used for: headlines, descriptions, social posts, scripts, culling, new topics
- 0:49 Asking it what it is, and how a business would use it
- 1:58 Why an incomplete follow-up question still gets a useful answer
- 2:25 What it says it can do for video, and the answer cutting off mid-sentence
- 3:10 The odd TikTok comments, and not wanting to answer them emotionally on behalf of the company
- 4:00 Pick a paragraph, post it, be done with it
- 4:20 One more refresh, and the model still will not answer
Links and resources
- The pillar this episode sits under: set up a workflow that works
- Chris Schwager on LinkedIn
Questions people ask about this episode
How do you create an AI generated video?
Not from this episode. Recorded in January 2023, it is a record of one video business putting ChatGPT to work, and every single output was text: headlines, descriptions, scripts, new topics and replies to social comments. No footage was generated at any point. The camera, the room, the lighting and the person in front of it were all still Chris's job.
Was the AI tool free to use?
In January 2023 it was. Chris names chat.openai.com on air, says it is costing him nothing, and calls the data he is getting back great. He also hits its limits twice in five minutes: one answer stops mid-sentence, and a refresh at the end of the episode still does not return a complete response.
What was AI used for in a video production business?
Writing, and nothing else. Chris lists it on air: more succinct headlines, descriptions, social media posts, scripts, abbreviating, culling, coming up with new topics and expanding on ideas. Two of those are the ones nobody plans for. Culling is shortening something already written without losing the point. Expanding is turning a one-line thought into something you can argue with.
Is the podcast itself AI generated?
No. Chris recorded episode 137 himself in January 2023 and opens it by calling AI his copywriter and scriptwriter, then telling Christine not to feel threatened. What was AI generated was the material he read out: the answers arriving on screen as he typed questions in live. The voice, the framing and the judgement about what to read are his.
Full transcript
Episode 137, recorded January 2023.
Read the full transcript
Chris Schwager: Here we go. AI generated podcast after 136 episodes. This is how we roll. My copywriter, my scriptwriter. Christine, do not be threatened. There will still be a use for you. This is early days in the process. AI is helping us to do is to make sure that we get things more accurate. So it becomes kind of more like the com specialist for the business, right?
[00:00:28] So this is, AI for us has been about, you know, writing more succinct headlines, descriptions, social media posts, scripts, you know, abbreviating, culling, different things, coming up with new topics, expanding on ideas. I mean, it is a range. Things. And now I've just typed into chat. G P T, what are you? And this is the response.
[00:00:49] I'm a large language model trained by open ai. I'm capable of understanding, responding to a wide range of questions and prompts. My main purpose is to assist users in generating human-like text based on the input provided to me. Right? How do businesses use it? Is my next. Businesses can use a large language model like me in a variety of ways, such as content generation, customer service, data analysis, and I'm actually taking out all of the stuff there.
[00:01:21] I'm just giving you the highlights, natural language processing. Virtual Assistance. Pretty interesting. Research and development businesses can use me for research and development in the field of natural language processing, natural language understanding, and generating machine learning and deep learning.
[00:01:36] These are a few examples, and the possibilities are constantly expanding as the technology develops. So what? And this is the good thing about chat j d, so I'm, you can see me typing away here if you're on the YouTube channel or on social media. You can see me typing straight in. I'm not gonna try and type an entire question again or, you know, to, to try and understand this in more detail.
[00:01:58] It's actually learning what I'm, where I'm going and it's looking at my previous correspondence here and gluing this together for me so I don't have to go, alright, oh, how does businesses use it for blah, blah, blah, blah. I'm gonna say to it something quite specific. I'm gonna say, what about. For videos, it's an incomplete question, but it's gonna see it and understand exactly what I'm talking about.
[00:02:25] Large language models like me can also be used in video industry in a variety of ways, such as one script, generation two, submitting and captioning. Three video des. and, sorry, I can't complete the response, , so you've gotta expect, I've spoken to so many people that are already on it. and using it and implementing it into their business.
[00:02:50] It's chat.open ai.com is the the application that I'm using, and it's free, so it's costing me nothing and I'm getting this great data. But in this particular instance, I'm sort of expecting that this could be some of the limitation early on of using the application because you can't get access because either there's too much, it's kind of getting overloaded.
[00:03:10] There's so many people kind of using it and trying it. But you know what? It's been unbelievable, you know? Talked about, uh, responding to social media posts. So I had a weird one couple of weird, oh, it's very , varying degrees of weird comments come through on, uh, TikTok, but instead of replying, you know, in some sort of emotional way attached to it, you know, on behalf of Rich Films, I don't wanna be doing that.
[00:03:36] I'm lying in bed, man. I wanna go to sleep. I feel compelled as a company to be replying to these people. So instead of trying to come up with the perfect response to their reply, I simply take their comment and it spits out several professionally written and really pro really, really good. I've just said, simply reply to this, and the responses have been great, and I.
[00:04:00] Pick a paragraph, put it back into social media and respond and be done with it. No energy, no brain space needed. And in the world of video where there's so many tasks already applied to the video production process, having uh, the ability to lighten this load around message development, I think is a huge advantage for.
[00:04:20] Uh, video producers moving forward. So check it out. If you can get onto it, it's chat.open ai.com. It is super cool tool. I'll see if it's gonna wake up now, and I'll do a quick refresh here. Your requester model is not complete. Okay, so that's it for this episode of Video, the podcast . As you can see, a little bit of improvement there, but, uh, some powerful staff in terms of content creation and delivering that message.
[00:04:48] That's all for this episode. Have a great week. See you next week.
AI will happily write the headline, the description and the awkward reply. Getting in front of the camera every week is still yours. Book a call and we will work out what is actually stopping the videos getting made.
