The AI Stack for Content Creators (aka the AI Tools I Actually Use)

0:00Today on the AI Breakdown, we're looking at a content creators AI stack. In other words, the AI tools that independent content creators can use to improve the speed and efficiency of their work. The AI Breakdown is a daily podcast and video about the most important news and discussions in AI. Go to breakdown.network for more information about our YouTube channel, our Discord, and our newsletter. Welcome back to the AI Breakdown. A couple of months ago, I did an episode called The AI Tools I Actually Use, or something like that, and people really responded. Now, I had wanted to do that because at the time, one of the big categories of content in places like Twitter and Facebook was people saying, Using chat GPT is just scratching the surface.

0:44Here are 18 ,000 tools that you need to know. I think that it served mostly to get people to feel overwhelmed and disempowered rather than what it was actually going for, which is to make people feel like there was a tool for absolutely everything that they could imagine. So, the goal of this video is not in any way to suggest that this is a comprehensive list of all the tools that content creators can use, but is just a set of tools that I find myself using in a relatively frequent manner, so maybe this will inspire something for your own content creation and help you get your own AI content creation stack up and running.

1:17Let's start in the realm of research. Now of course, basically any LLM, especially ones who are connected to the internet, are going to help you with research, but I wanted to pay specific attention to perplexity. Perplexity is increasingly the research tool of choice for people who I know who are familiar with lots of the different AI options, and there are a few reasons for that. Let's take a look at the example that I'm using throughout this video by asking Perplexity, what is OpenAI announcing at Dev Day on November 6th? A few things about Perplexity that are worth noting. First of all, right up front, it shows you the sources that it draws from, and that happens before it gives you its answer.

1:51Now, this was sort of a hard question, because I didn't just ask for a description of what Dev Day was, but asked what they were actually announcing, the trick of which, of course, is that they haven't said what they're announcing. Perplexity comes back. OpenAI is hosting its first-ever developer conference, OpenAI Dev Day, on November 6, 2023 in San Francisco. While the specifics of the announcement are not detailed, OpenAI has indicated that they will be previewing new tools at the event. However, it has been confirmed that the announcement of GPT-5, the presumed name of OpenAI's next flagship generative AI model, is unlikely, as OpenAI CEO Sam Altman confirmed in April that OpenAI wasn't training GPT-5 and wouldn't for some time.

2:26Now from there, it begins to speculate about what might be revealed. Quote, There might be updates on OpenAI's plans for Global Illumination, an AI design studio that it acquired in August, and an update on the availability of GPT-4's image understanding capabilities. There are also reports that OpenAI plans to unveil new, possibly cheaper vision-capable developer tools at the conference. Now for those of you who are listening rather than watching to this, one of the things of note is that after almost every sentence, certainly after every section, there are little footnotes which show you which source different information came from.

2:56So even if it was just footnotes and summarization, it would already be a really useful research tool. But one of the other things that makes Perplexity really cool is the way in which it prompts further or related questions. For example, it gives me the option to click, What are the new tools that OpenAI is planning to announce at DevDay? How can developers register for in-person attendance at OpenAI DevDay? What is the agenda for OpenAI DevDay? Let's try the agenda. Now in this case, it actually came back and asked me for further information. Let's do it suggestion, keynote topics. Now, the answer here is pretty vague because, again, this information hasn't really been premiered, but you can see how this related question type of feature is really good for getting further information and for digging into a topic quickly.

3:35Let's move on now to another use for AI in the content creation process, which is scripting. Now, it should be noted that when it comes to these videos and these podcasts that you're hearing on the AI breakdown, I am completely unscripted. Every day I figure out what I want to talk about. I pull up the web windows, and I just let it rip. But of course, that's not the way that most people create videos and podcasts. And one of the things that these new LLMs are really good at is writing, which of course, with a little help, can turn into scripting. So we selected the web browsing model of GPT-4, and first I asked, what do you know about the OpenAI Dev Day happening on November 6th?

4:08It gave a little bit of information, which for me was really just about making sure that it knew what the context was for the thing that I was about to ask it next. I said, could you please write an intro script of 15 to 20 seconds for a video all about the event. Now this actually was quite the fail. Its first suggestion began, Welcome to OpenAI Dev Day 2023. Unfolding this November 6th in the heart of San Francisco, this inaugural event ushers in anew. Now at that point I knew that it was thinking that I kind of wanted a hypey video or that I wasn't thinking journalistically. And just to confirm that I clicked to generate and still had something similar.

4:39Get ready to step into the future of artificial intelligence at OpenAI Dev Day 2023. As you can see, this type of scripting was almost more like a promotional video than it was like what I was looking for, which was the intro to a video that was more journalistic in nature, or more news analysis. So I stopped it and added this clarification. I'm a podcaster who covers news about the AI space. I'm doing a video about OpenAI's Dev Day. Can you please write a script for a 15-20 second intro about the event that focuses on what's interesting about it or the questions people are wondering about it?

5:08Please don't seem like you're shilling the event or hyping it up too much. This one, it did much better. ChatGPT responds, Absolutely. Here's a balanced intro script for your video. OpenAI's inaugural Dev Day on November 6th unveils a space for global developers to converge and cross-pollinate ideas. As attendees get a first glimpse of new tools, the AI community is abuzz with questions. What innovations will be revealed? How will OpenAI's latest strides shape the development landscape? Join us as we delve into the unfolding narrative of OpenAI's Dev Day 2023. Okay, so we are going to use that script for a couple other demonstrations throughout this video.

5:42The first of those is speech synthesis. If you are a regular AI Breakdown listener, you will know that the time that I use speech synthesis most often is when I'm doing one of the long reads episodes where there's some big 10 or 15 or 20 minute piece. I will often have AI me read it instead of myself. But let's do a quick listen to hear how 11 Labs synthesizes my voice, which I've trained specifically for 11 Labs, doing that intro that ChatCPT just wrote. OpenAI's inaugural Dev Day on November 6th unveils a space for global developers to converge and cross-pollinate ideas. As attendees get a first glimpse of new tools, the AI community is abuzz with questions.

6:17What innovations will be revealed? How will OpenAI's latest strides shape the development landscape? Join us as we delve into the unfolding narrative of OpenAI's Dev Day 2023. Now, as you can tell, it has perhaps a little bit less of the vim and vigor that I might if I was doing the intro to one of my own videos, but it's certainly not wooden and it's not dead and it certainly sounds a lot like me. Now I did want to throw out that there is another tool in the same category that's actually kind of a more comprehensive tool. The editor that I use for all my videos and podcasts is Descript. Descript allows you to edit videos by selecting text and it has a number of different AI features including the ability to automatically cut out all the filler words like ums and ahs.

7:00It has an overdub feature where you can actually type in words and have it speak in your voice in the right intonation so you can fix things in post-production and it also now has AI-assisted writing that you can then have a voice that you've trained read as a script as well. So just for giggles, even though it doesn't necessarily have the context and I'm not sure if it's wired to the internet, I ask it to please write a 50-word script introducing a video about OpenAI's Dev Day developer event on November 6th. Let's see what it comes back with. It writes, Welcome to OpenAI's Dev Day, the must-attend developer event happening on November 6th.

7:30Join us for a day of deep dives, technical talks, and interactive sessions with experts. Discover the latest advancements in AI research, explore cutting-edge technologies, and connect with a vibrant community of developers from around the world. Now, obviously, this is doing the promotional video thing, but just because I'm doing it as a demonstration, let's add it to the script, and then click Add a Speaker and see how it sounds. Welcome to OpenEye's Dev Day, the must-attend developer event happening on November 6th. Join us for a day of deep dives, technical talks, and interactive sessions with experts.

8:00Discover the latest advancements in AI research, explore cutting-edge technologies, and connect with a vibrant community of developers from around the world. Don't miss out on this exciting opportunity to learn, network, and shape the future of AI. Save the date and mark your calendars for OpenEye's Dev Day on November 6th. So the way I would describe this is that it would be very impressive if it was about six to nine months ago. But now, obviously, it doesn't really hold a candle to 11 labs, although there are options that I haven't taken to further train and fine-tune a voice model in Descript, so it could be more MyAction than Descript itself.

8:35However, when it comes to the use for patching videos, it's an unbelievably powerful integration that makes editing a totally different type of experience. But what if I don't just want audio? What if I actually want video, and for whatever reason, I don't have the time to get on camera? Well, as HeyGen's landing page puts it, No camera, no crew, no problem. So this is what HeyGen looks like inside. I've trained one of these instant avatars on myself and then fine-tuned it, and so let's get a preview of what it looks like before we see the video with this specific script applied. Hey, Nathaniel Whittemore, your instant avatar is ready.

9:09Feel free to create videos with it. Also, click the feedback button to share what you think. Hope you enjoy. Now, one of the big reflections that I had upon seeing this instant avatar video is that I probably bounce around too much when I'm filming, because the AI has clearly picked up and mimicked that action in a way that I find fairly distracting, but frankly, that is a notch in the AI's favor, as it actually picked up on a distinct tick that I have when I shoot my videos. Now, when it comes to creating a video that has this specific script, you click create new video, you add speech, or I've copy-pasted the ChatGPT script, and then I've selected the avatar.

9:41I'm going to press submit, and then we wait for it to be ready. OpenAI's inaugural Dev Day on November 6th unveils a space for global developers to converge and cross-pollinate ideas. As attendees get a first glimpse of new tools, the AI community is abuzz with questions. What innovations will be revealed? How will OpenAI's latest strides shape the development landscape? Join us as we delve into the unfolding narrative of OpenAI's Dev Day 2023. So as you can see from that video, this is far from perfect. There's a lot to be said for just getting on camera and actually doing it yourself. But man, just the availability of something like this with an avatar that you can train on yourself opens up so many new possibilities in terms of the speed of content creation, the precision, and frankly, the cost.

10:31Although yes, there is a cost to using HeyGen. It's nothing compared to the time that an editor takes. Okay, so we've got our video now. Let's make some key art, such as for a YouTube thumbnail. For any given image that I'm working on, there are two services that I use most often. Those are DALI 3 and MidJourney. Now, MidJourney for a very long time was my go-to. The new capacity that DALI 3 adds is the ability to actually have words inserted, which can make a huge difference. You can see here that I've asked, please make a YouTube thumbnail, 16x9, for a video about OpenAI's Dev Day developer event on November 6th in San Francisco.

11:04It created two, one that has a Golden Gate Bridge theme, and another that has a sort of cool computery illustration view of San Francisco with some generic tech accents to make it feel modern. Now these are fine, but if I wanted something perhaps a little bit more artistic, mid-journey might be a better call. The prompt I used for this was poster image for an artificial intelligence event for developers in San Francisco. Now these aren't perfect enough, I was doing this for real. I would certainly be refining them and experimenting with different prompts to try to get a little bit more of what I wanted.

11:32But stylistically, number two is close enough to be going on, and so I wanted to bring it to the last tool that I use very regularly, which is Canva. Now Canva has tons of non-AI features. It's just a good place to lay out images, be they thumbnails or something else. But they also have a whole new set of AI tools, one of which is called Magic Switch. Part of Magic Switch is a magic resize. So for example, I could turn this into a square thumbnail just with a single click. Let's do that and see what comes back. We click copy and resize, then open it up. Boom, sure enough, it's there. Now of course, with something this simple, it wouldn't have been all that hard to just do this manually.

12:06But sometimes the point of these AI tools isn't to be totally revolutionary. It's to save seconds or minutes that you could use for other purposes. But now let's try a different capacity of Magic Switch, which is translate. Let's translate it to Spanish. This sort of linguistic translation is something that I think is going to be hugely significant, and in fact gets to the last category, which is a toolset that I haven't integrated yet, but which seems like it's inevitably just around the corner. That is the crazy world of AI-powered dubbing. Now, a lot of the big platforms have started to announce dubbing features, including YouTube and Spotify, but a number of startups like Wondercraft are also offering this as well.

12:42My strong instinct is that within a year or two, the default will be for a lot of the content that we create to be automatically translated into at least another half dozen languages or so, which if that actually happens, will break down global linguistic barriers in a huge, huge way. And so friends, that is where we will wrap. Like I said, this is far from a comprehensive list of AI tools for content creators, but it is one content creators AI stack. I hope, of course, that you found this useful and that it inspired some ideas for you to go off and use on your own. I appreciate you listening or watching as always.

13:14And until next time, peace.