These are all AI tools that have been out for a while, like a few months or longer. They each sounded cool, but I put off trying for whatever reason. Then when I eventually did, I was blown away and regretted not trying them sooner.
This list goes all over the place from research, to music, to open source databases, to content creation, but they're all super useful and most of them are free. Napkin AI turns text into visuals that are perfect to use in video content. newsletters, blogs, presentations, tweets. It's currently completely free while it's in beta, so that will change at some point, but it's really easy to jump in and use.
Just click new napkin, then either draft with AI, or first I'll start from a blank napkin and I can input my own information. Here's a post about the most used AI tools. I'll copy that, then paste it in here, and I'll see what it can do with that.
Just click on the little lightning icon and it will come back with that data incorporated into diagrams, flowcharts, and infographics which are all fully editable. This one looks pretty cool and each of these have additional styles within them. I like this one so I'll select that. Then I can change any of the text.
I can switch out colors however I want to. It's easy as that. Then you can download it.
I'll do another one really quick. New napkin. This time I'll draft with AI.
How about all the steps involved in training in LLM. It's all written out. That looks good.
I'll select all the text to create a visual from. There's a ton of good options in here. Now I want to show an example of how I use this tool for videos. I could take this spiral one and blur out the future steps, and as I go through, I'd be zoomed in, then unblur as I move on to explaining each step.
Or this one. I'd zoom in on each area as I'm talking about it. Same with the timeline one. That's pretty straightforward. Graphics like this are really time consuming to create from scratch.
I'll do one more example to get some more varied results. things most likely to cause the AI apocalypse in order of likelihood. All right, that looks like a solid list.
This time I'll use just the first one to generate from. I like this fishbone one a lot. That seems very fitting.
Now I can select all of these and try again. I'll just do the top five instead of all 10. And there's a bunch of good ones in here. I like this chain with all the links, so I'll check out the other styles. I prefer the ones with the icons in here and all the colors. This one is perfect for a video.
I could just add some zooms between these and have a nice visual to make it more engaging. As you generate a few of these, you'll notice it repeats the templates, but it does have a lot of them, especially when you factor in the different styles for each. So I've found this to be a really useful tool that I'll be incorporating into videos.
And there's tons of other use cases like visuals for a newsletter or blog or a presentation or just a quick social media post. Google's Notebook LM is another free one. There's been a ton of talk about it and I tested it a while back, but it took me a while to actually start incorporating it.
I'll show a couple of my favorite use cases. I'll create new. Here's the options you can upload. Slides, PDFs, websites, or I'll do a YouTube video. This is Andre Karpathy's presentation about how LLMs are trained.
It's really good, but I don't want to watch the whole thing again. I just want to revisit the information. He also posted the slides, so I'll upload those as well.
Now it generates a summary. Then there's these options to choose from. I'll go with study guide. Using the information in that YouTube video and the slides, it created a short quiz with an answer key, some essay questions, and a glossary of key terms.
That's really useful. And I can chat and ask questions directly. Explain the fine-tuning stage. It recaps that for me. Then for all the sources, when you hover, you can read the text it came from.
Or if you click, it will pull it up in the document so you can read the context and double-check the accuracy. Then click Notebook Guide to get back. And I'll generate a podcast.
This is the feature that's gone the most viral because of how authentic these sound. It takes a little bit, so I'll come back when it's done. Let's give it a listen. This one's actually a bit longer.
They're usually only five to ten minutes. All right, everyone, get ready, because today we're diving deep into large language models. Hell, LMs. Yeah.
I'll skip forward. They usually take a little bit to get into the substance, so pretty accurate for a podcast. It's like Lama 270B are basically just two files.
That's it. Really? That seems, I don't know, almost too easy.
It's true. One file is just packed with numbers. We call them parameters. Think of it as the model's brain.
And this brain can be huge, like 140 gigabytes for Lama 2. Wow. Then the second file, it's just a short program, basically. It tells a computer how to use those parameters.
So hold on. You're saying I could run this advanced model offline, like on my laptop? That's kind of crazy.
These always come back amazing. I'm still blown away every time that this is possible. Another thing you can do is you can upload a bunch of studies on a topic and ask about them.
That's super useful. There's been a lot of talk about this one recently, and you can check and uncheck sources to speak with each individually or together, however you want. I can ask about all these studies as a whole, which included a few meta-analyses. Then I could click through the sources and go back and forth.
I'm sure you get the idea. It's the best UI I've seen for doing this. And these can take in a lot of information.
For example, I can upload a full book. The Omnivore's Dilemma is a good one. I have a physical copy and- notes I took on it, but I can't ask questions or stay as organized that way.
But now I can chat with the book, create a study guide, and listen to a podcast about it to help retain the information. And you can customize what they talk about with these too, so I'll narrow it down to a more specific section of the book. Yeah. So we're talking about regenerative farming. Okay.
And to really dig into this, we're going to be focusing on a pretty legendary place. Hit me with it. Polyface Farm. Heard of it.
Yeah. It's pretty amazing. So this is amazing for learning and studying.
You can reinforce the information while doing the dishes or driving to work. And another more unique way to use this is if you create videos like I do or need to give a presentation or write about a topic, you can upload the information you're working from and generate a podcast about it and then get... ideas for how you could present it yourself. These have good storytelling and unique angles, so it's just a good way to brainstorm. There's actually a new feature they rolled out since I recorded this called interactive mode.
You can basically call into the podcast and ask questions so you can help steer the conversation as it's happening. Something that's happened to me quite a few times is when I open up cloud or chat GPT and I know it could help me with something I'm working on, but I can't think of the right question to ask. Well, if you're a marketer, entrepreneur, or creator, HubSpot has taken. all of that uncertainty out with a list of 1,000 prompts that you can download for free.
1,000 prompts sounds like a lot, because it is, but it's broken down into different categories to make it more efficient to find the right prompt. There's marketing strategy, copywriting, social media strategy, competitor analysis, productivity, and a bunch more. There's 18 different categories.
You can copy and paste each prompt, then there's usually a couple fields to customize to your specific industry or goals. It's a great way to craft strategies, generate ideas, and save time. and is tailored to give actionable responses. It's like a personal productivity assistant.
Again, if you want to grab that, it is a completely free download from the link in the description. And thank you to HubSpot for sponsoring this video. Pinocchio is a tool that completely simplifies the process of downloading open-source AI tools.
It's like Steam for AI apps. Usually setting up an open-source model involves cloning a GitHub repository, installing dependencies, downloading models, configuring everything. It's a multiple-step process that's pretty involved. But with...
Pinocchio, it's a one-click install. If you have a capable computer that's able to run these types of models, Pinocchio is a massive time saver. And every aspect of this is free.
So to get it, go to pinocchio.computer. Then it's just like installing any application. Mine's already installed. This is what it looks like once it's open.
The Discover tab has all the apps to look through. I stick to verified scripts. The community scripts aren't verified to be working. They just show up when a developer tags Pinocchio on GitHub. And either type in the tool you want or scroll through until you find one.
Then click on it to open it up. Then click download and let it download for a while. Some of these are pretty large files depending on the tool. Once it's done, you can find it on your home tab.
You can open it up, then they'll each look slightly different, but there should be a button that says launch or something similar. I've got Live Portrait up and running, and here's an example I used for a previous video. I filmed myself singing the words to a song I got from Suno. Then I transferred my facial movements onto a video I generated in Kling.
There's a bunch of other really fun tools like that that you can download with just a click. So I'm sure you'll find something useful in here. It is weird to trust something like this, but it's been out for a long time with a lot of people using it.
The creator Cocktail Peanut has a great reputation. He's on X and you can join the Discord to get updates and chat. So it is safe, but I'll still throw out the caveat to do your own research anytime you use something like this.
And there's some tools that won't be on Pinocchio, especially the smaller one-off tools. My favorite cloud solution is Replicate. It's also the solution if you don't have a computer capable of running models on it. Or maybe you just want to test something before going through the process of downloading it locally.
Replicate is a pay-by-usage model. You just put a card on file, then get billed based on what you use. There's a huge database of tools in here for upscaling, text-to-3D, text-to-video, text-to-speech, image restoration, music generation. The list goes on and on. I like Replicate a lot because sometimes you just need to use a tool a couple times, but most AI tools only offer a monthly subscription.
A lot of the open source stuff is just... awesome. The best ones usually get integrated into platforms with a nicer UI, but you can use them all more cost-effectively here. Like the new Flux image editing tools for inpainting, retexturing, and restyling.
Or they have minimax on here. Maybe you just want to use it a little bit, instead of paying the monthly subscription on their site. You can just use what you need here. There's tons and tons of other tools to experiment with. One of those that I just demoed recently using Replicate, but deserves its own spot on this list, is training a Flux Laura for consistent characters.
I trained it on my own face, and it's insane how good it is. And now that I have it trained, I can use it at any time for thumbnails or to turn into random B-roll. It costs a few dollars to run the original training, but after that, it's only a couple cents to generate images with your face whenever you want.
Just a quick run through of the process. Search for Austris, then click this one that says FluxDev Laura Trainer. First, write a name for where the model will be stored. Then click Create New Model, and I'll keep mine as private.
For the training, you need 10 or more photos. I mostly used my thumbnails. Compress the images into a zip file, then drag that into the box to upload. Next, create a trigger word, which you'll refer to in your prompts to trigger it to use this character. Now increase the LoRaRank to 32, which will help to train on complex features.
Everything else is good, then just click Create Training. And these take about 20 minutes to finish. Once it's finished, you can access this and any other models you train at any time from your dashboard. Then type in a prompt. Kevin was my trigger word, so...
Kevin is a Viking charging into battle. I'll use a 16 by 9 aspect ratio, switch the number of outputs to 4, then scroll all the way down and switch to PNG, then run. It generates really fast and the face is basically perfect.
It's really nice to just have one of these trained up to use at any time. If you have a YouTube channel, this is one of my favorite tools I've been using a lot lately for brainstorming title and thumbnail ideas and choosing video topics. It's called Spotter Studio.
I'll open up a new project, then type in a title. I've been thinking about making a video on all the recent advancements in AI and robotics. I can have it rephrase. That's helpful.
Ask it to shorten it or explode is a great one. We'll take the idea in all different directions. Then you can find your favorite and have all those options open up again. Then just reiterate until you find something you like. And back over here, I can brainstorm a thumbnail idea too.
I'll just have it generate based on that title. And since I've uploaded my face, it will use similar faces across these. There's options for more realistic or more sketch based, and a slider to have it match my own thumbnail style more closely or others across YouTube.
The outliers tab is another really helpful section. It finds the best performing videos across YouTube. What's even more helpful, since I have my channel linked, there's a tab for channels my audience also watched. So I'll switch up the settings a little bit to find the highest performing videos, and it finds the outliers relative to that channel's other videos. So this video has done eight times better than this channel's average.
That's how you know it's a topic that people are finding interesting or valuable. I used to do outlier research manually, just going to different channels one by one, but Spotter makes it super easy. And you can click brainstorm to ideate from there directly and try to find a unique perspective on that topic. All right, they didn't sponsor this video, but full disclosure, they are a sponsor of the channel and sponsored my last video.
But it's a super helpful tool that I actually use. But because they sponsored that one, they did give me a link that will give you your first 60 days free instead of 30 if you... decide to sign up through that. Guide is such a useful tool if you ever need to show someone how to do something on a computer for any reason. Whether it's someone at work you need to show a process to, maybe a full onboarding system to train new people, a blog, a course, whatever it may be, guide can save a lot of time.
I always want to read it as Guide from the spelling, but it's probably guide. You have to download the extension and then ask it to start the process, and then you just do the task. So I'll walk through the steps of training a Fluxalora again just really quick. Then I stop recording and it will create a full slideshow and video with a voiceover of the step by step process highlighting where to click and everything you did with a written explanation alongside. Depending on the process, it may not be perfect the first time, but you can edit anything.
I'll take out a few of the steps, then change the wording on some of them. It's really quick to do. Then you can click edit video to change the text on the screen and the arrows.
You can switch to a different voice if you want. It's all really intuitive. Then once it's done, you can share it and what the receiver will see looks like this. So there's the step-by-step explanation and images, then the video up top.
This guide will walk you through training of Fluxalora. Begin by clicking on get started. Proceed by selecting a model.
Choose a name for your destination. This is really helpful and you can create 25 guides for free, which may be all you need, or it's at least a good trial to see if you like it enough to buy. This next one isn't the tool itself, but a feature in it.
So, Suno is amazing for generating music and has been for a while, but I recently tried their feature for uploading audio and extending it. The results were amazing and it's a whole different way to interact with AI music. So, here's a bass line.
Then I took that and uploaded it in Suno. I had it generate some lyrics using the prompt, a robot learns to play funk, then asked it to extend. And here's what it came back with. And they also have the cover song feature where I can take that same bass line and ask it to create a cover song. It uses my upload as inspiration so the bass sounds similar but not as exact.
I love that feature. It has been out for a while, but I just tried it recently. Using AI as a creative partner instead of just typing a prompt is a lot more fun. I don't like having AI just do everything. That was just a quick example.
There's a lot of possibilities, and I hope to see them expand more in this direction towards creative control and collaborative output. Those are some tools that have been extremely useful for me and saved me a ton of time. Hopefully that's helpful. If you want a video focused solely on AI video and animation tools, check out this one right here.
And make sure to check out futurepedia.io to find the best AI tool for any use case. Browse a curated list of AI tutorials, stay up to date with AI innovations, and sign up for the free newsletter. Thank you so much for watching.
I'll see you in the next one.