AI Video Production Workflow — Step-by-Step Guide & Transcript

[00:00] Thanks to AI, I've cut down my video production time to just a couple hours, and I want to show you how you can do the same. So if you're ready to save some time and pump out more content, let's dive in. Now my workflow is pretty straightforward. I'm only using a couple of AI tools, and the best part, they're all free.

[00:16] So you can just copy my system without spending a single penny. For every step mode to show you, I've got a stopwatch ready to let you know how long each one takes. It all kicks off with ideas and research. Now, you might expect me to whip out ChadGPT and drop a basic front like, suggest video ideas.

[00:33] But if you're real, everyone knows ChadGPT by now. And to be honest, the ideas he spits out aren't exactly top tier. No offense to GPT fans, but if you're a YouTuber like me, it doesn't quite hit the mark. So instead of relying on it, I do my own digging.

[00:48] Checking out competitors and keeping up with trends. Usually, I already have a rough vision of what I want to talk about. I'll hop on YouTube, find a few similar videos, and start correcting their ranks. Once I have them, I dump those ranks into a free tool called NoGPT.

[01:04] Now, let me tell you, NoGPT is a game changer. And no, not sponsored by NoGPT in any way. What it does, it drafts a transcription, summary, and mind map for every video. That mind map is totally useless for me.

[01:18] Skip it if you're following along. The transcription is accurate and has time codes, but again, not super helpful for my process. What I do care about is the summary tab. This is where the goal is. Give you highlights, key insights, and a super short one-sentence summary,

[01:34] perfect for figuring out what the video is about without actually watching the whole thing. I either copy these notes somewhere or jot them down on my iPad or even just good old paper. But I don't stop there. The fourth tab in Node.GPT is the AI chat.

[01:49] and I use it to talk with the video's content. Yeah, you can literally ask questions like, does the video hint at anything? Normally, I would have to sit through the entire video to catch those little details,

[02:01] but by chatting with it, I get a quicker understanding of its angle. This helps me find ideas that weren't explored or figure out how to approach the topic differently. I repeat this for three to five videos, and by the end, I've got a crystal clear idea

[02:16] of what my video will be about. For instance, if everyone's making videos about why GDP 5 might flop, I might split the script and talk about why it's bound to succeed. Just an example, but you get the point.

[02:28] So after I finish watching and researching videos, I usually dive into a couple of web articles. Well, dive might be stretch-nit, I skim the summary so I have the real, and I use the same no-GPT tool to summarize articles.

[02:41] Most tools out there, yeah, they make you copy-paste the text yourself, which is such a pain. Why can't they just let you drop in a link, right? Summarizing an article works exactly like summarizing a video for me.

[02:53] I check the summary, ask AI follow-up questions, and take notes. Videos, they are for ideas, and articles, that's where I get the hard cuts, the data I need. And you know what? In 15 minutes, I'm done.

[03:05] Normally, this would take an hour, easy. I guess the only reason it doesn't take 5 minutes is that I'm kind of serious about a leaker's game. So I can edit in for a while and yeah, can get real tedious. But I found a way to speed things up.

[03:17] I use Vee, the sponsor of today's video. Specifically, their Magic Hood feature. Absolute lifesaver. You know those awkward tosses, the uhs and ums we all do? Yeah, Magic Hood just wipes those out with no effort.

[03:31] One click and boom. My video is cleaner, tighter, and way easier to work with. Changes the world and creates innovations that affect millions of lives. And V also has this eye correction feature, sometimes I'm not staring directly into the camera, don't judge. Happens.

[03:47] And instead of redoing the whole thing I just hit a button and V fixes it It probably some AI wizardry happening frame by frame but it so smooth you wouldn even know I cheated Love that Of course the basics are solid too Noise removal yup They got this algorithm that wipes out background noise like it nothing and it super fast

[04:08] You've decided to record a YouTube video, but... You've decided to record a YouTube video, but... Plus, the subtitles on point. They're accurate, customizable, and you can even throw in animations or emojis. Yeah, it's that extra.

[04:20] But honestly, this is just fresh and the surface. These get all sorts of AI-driven tools, like translating audio, turning text into video, creating AI avatars. It's insane.

[04:32] If you're serious about video editing and want to save a ton of time, Bede is where it's at. Seriously, go check it out. The link's down below. When I'm done with my research and I know exactly what I want,

[04:44] I head over to ChadGPT. But I don't just throw everything at it than expected to write a whole script in one go. And that's something you need to keep in mind. If you give CAD GPT a huge test up front,

[04:56] it's probably gonna start hallucinating, repeating itself, or just rephrasing what you gave it over and over. The result, a nasty, inconsistent script. Instead, I take a different approach, custom GPTs.

[05:09] I've got two of them. The first one is designed just for outlining the rough structure of the video, giving a clear flow and direction. you're curious about how to create your own custom gpt make sure to check out our recent

[05:21] TED gpt guide I break down everything you can do with it so for this first gpt I give it a simple prompt idea plus column plus the idea itself and at the end of the prompt I usually add

[05:35] use the web for information just to make sure it's good I've got web access enabled by default in the GDT settings, but it doesn't always go as planned. Even with a custom GDT trained on my exact preferences, ChaiGDT can sometimes just do its own thing. When that happens, I just tweak

[05:52] the prompt slightly. For example, I might ditch the use the web request. Honestly, you never know which part of the prompt is going to make ChaiGDT go off track. Luckily, this time it worked fine.

[06:04] Here's how it's set up. ChaiGDT splits the script into sections. Each section starts with a little intro to grab attention and connected to the previous part. Then give the list of options that should be covered in that section. The bullet points though are usually pretty basic,

[06:20] no matter how I prom them in the GDT settings. So I've gotten into the habit of asking Chad GDT again to use the web and fill in those bullet points with more detailed information. Add more

[06:32] depth, more useful insights, and sometimes even links to sources. This outline pretty much becomes the backbone of my videos. Now the second custom GPT, I don't use it as often, but

[06:44] it's the one that actually writes the script. I train it by uploading a bunch of our alt scripts so it knows our style, our tone, and the vocabulary we use. I set it up so I can just copy the outline from the first GPT, paste it in, and boom, the

[06:59] full script. Even here, I have to tweak things as I go. By default, ChatGPT is set to give shorter answers so I usually have to follow up with an extra prompt asking it to expand. That's the secret sauce to get in the right length for my scripts. What you need to remember when

[07:14] working with CADGPT is that no matter how well you set your custom GPTs, it will still mess up. You need to stay on top of things at all times, controlling every move. In my case, I only have to do minor adjustments and that's not black magic, it's just prompt engineering. And by the way,

[07:31] you can learn all my tricks in our beginner guide to ChadGBT. It's completely free. Now, I want you to take this script right in part with a grain of salt. It's been another 10 minutes and the script is mostly done. It's not perfect in any way,

[07:45] but with some of my narration magic I can make it work Though sometimes the results ChadGBT gives are far from perfect and straight up refuses to fix it These cases I either start a new chat with this system GDT

[07:59] or head over to Claude and try to make it work there. With Claude, it works pretty much like you would expect. There's one catch. It can't access the web. So yeah, it can pump out scripts, but they tend to lack the latest, most relevant info.

[08:13] And for some scripts, that's no biggie. For others, well, that's when I either have to jump in and add stuff manually or switch over to chat.gbt. But honestly, Quad is still pretty solid at cranking out scripts.

[08:26] You can even upload files to it, so if you had a script done and wanted to use it as a reference, boom, easy. Just know that there's no such thing as custom GBTs with Quad. So every time you chat with it, it either helps and remembers what you did before or starts fresh.

[08:42] Aside from that, it's pretty fast. Took me like three minutes to get a script done. When running this channel, I tested a ton of AI tools. Not all of them make it into the video, though. But if you stand out, those end up on my website, aimaster.me.

[08:58] I hand-test everything on there. So trust me, only the good stuff makes the good. And yeah, the list of tools is growing. As we test new ones, they will get added. So if you are into AI, go check out the site.

[09:10] Definitely a good starting point. So far, I've got my script ready to roll, and I'm about to start filming. No AI here, just me, my trusty camera, my, like, a bit of editing magic, and boom, we're done.

[09:23] The stopwatch says about two hours. Now, before you start expecting me to gush over some AI that edits my videos for me, let me stop you right there. That's not happening. Sure, there are AI tools out there that can edit video, do a pretty solid job,

[09:38] But for now, I prefer my videos edited by real humans. It's just how I like it. Same thing with the thumbnails. I'm not generating any fake images of myself. I just grab my camera and take a couple of good old photos.

[09:52] But here's where AI does come back into play. Even with a great camera, sometimes I like to give my photos a little boost. Make them pop, you know? And for that, I use CREA AI. Super simple.

[10:04] No weird tricks. I just upload my photo, select the upscale option, and that's it. Check this out. Now, full disclosure, sometimes the results can get a little, you know, interesting. AI tries to predict what every picture should be, and occasionally it misses, especially with my glasses.

[10:21] But, hey, the beard, the hair, spot on. Heads up, though, if your original photo isn't high-res, you might end up looking like someone else. Even with my good quality shots, AI has somehow altered my eyes before.

[10:35] Like, I can't tell it's me if I squint, but yeah, it all depends on the quality of your original pick. And hey, you don't have to use Fria AI. There are loads of AI tools for this, and we've covered 20 of them on the channel.

[10:49] Right now, the top watch does 2 minutes. Well, not bad, right? If you've been following the channel, you probably remember Magnific AI. I once crowned it the best AI uptailer that hasn't changed. It's still top tier.

[11:01] It's a paid app, but worth every cent. Today, though, we're keeping things free, so check out AImaster.me for more awesome tools like that. Once my photo is looking sharp, I jump into the thumbnail, edit it in like usual.

[11:13] Then I upload the video. No AI needed here either. But that's because I've been doing this for a while. If you're new to this, ChaiDBT can definitely help you out with things like tags. Also, descriptions, all bad.

[11:27] It won't nail it 100%, but it's more than good enough. So if I do a quick bit of math, the whole process from start to finish took about 2 hours and 30 minutes. If I was done everything manually easily another 2 hours packed on Nikki saying my content pipeline ends here so hold on We just getting started So from here there are two routes I usually go Route one

[11:52] making a bunch of shorts or reels. Route two, turning the video into a full-on text article. Now, I've already shown you a bunch of tools to create short clips from longer videos, so I'm not going to repeat myself.

[12:04] It's literally as simple as paste a link, wait in a bed, and boom, done. Easy peasy. I'm not heading into shorts or reels right now, but what I am doing is trying to add a blog section to the website.

[12:16] Yeah, I will be posting articles about AI there. I'm not a fan of writing my own copies, so what do I do? I take the audio from my edited video, transcribe it using VStream, and bam, instant text. VStream's awesome because it doesn't limit me on audio length. A lot

[12:33] of tools hide this feature behind a paywall, but not VStream. It's a huge win. All I do is upload my audio, hit a button, and in about a minute, for a 13-minute track, I've got my transcription. No need to sign up, no hassle, it just works. I usually do this when I improvise

[12:49] a lot during sonings, but it works just as well if you're using a script like something from Chad GPT or Claude. Well, next step, I take that transcription, toss it over to Claude,

[13:01] and ask it to turn the text into an article. Pro tip, ask Claude to suggest images that could fit the article. It won't generate them, but at least you will get a solid list to work with later. And of course, don't forget about SEO keywords. It's a blog post, right? Claude does a great job with

[13:18] that. But if I really want to make sure I'm covered, I will ask for those SEO keywords again, just in case. Trust me, SEO is everything when it comes to blogging. All right, we're almost there. Just one final thing to do. Grab a few images from the blog post. And don't worry, no surprises here.

[13:34] I've got three solid options and get to pick which one you like the best. Option one, chat GPT and Dolly. Super easy and hands down one of my favorite options. I've got a detailed guide on this, but in short, Dolly does exactly what I ask.

[13:50] I can copy an idea straight from Flod, paste it, and boom, an image. And if I need tweaks, just talk to it like I'm chatting with a real person. The downside, though, it's not totally free. You can generate a few images for free, but if you want to use it a lot, you will need a subscription.

[14:07] Still, it's the easiest by far, so I highly recommend it. Option two, Midjourney. I fully believe this one is my least favorite. The whole Discord chat van, not the most convenient, and the prompts, you've got to be super specific to get it right, which can be a hassle.

[14:24] But the results are consistently awesome. Quality, resolution, style, Midjourney nails it. So, if you don't mind the learning curve, you will get some great stuff. And, hey, if you need help, we've got a full guide on it for free.

[14:38] Option three, Adobe Express. This might be the best option if you're looking for something free and feature-packed. It's not just about generating images. It's got in-painting, background removal, resizing.

[14:50] Heads up, though, the interface is a bit chaotic. Everything strands into one editor, and it's easy to get lost, but the image generation is solid. You can pick a style, upload a reference, and even create collages.

[15:02] So, for versatility, Dolby Express is a win. Now, after writing that blog post with Claude and generating the images, the whole thing took me about 16 minutes and another 15 minutes for the shorts,

[15:15] and in about three hours, I got four YouTube videos, some shorts and reels, and at least one awesome blog post. And if I ever want to take a break from shooting videos myself, I could just create my own avatar with Heigen.

[15:28] That AI could basically host the channel for me. If you're interested in seeing a full guide on that, let me know in the comments. Thanks for hanging out, and I'll catch you in the next one. Peace.