Transcription
There are constantly new AI tools coming out. I test as many as possible for this channel, but there's very few of those that I find valuable enough to actually keep in my routine and workflows. Usually, that's tools that save me a lot of time. So, most of these are around content creation, productivity, and business.
First off is Whisper Flow. This is speech-to-text in every app on your phone and computer, but not like the standard dictation features. I was actually pretty resistant to using speech to text for a long time, but using Whisper Flow actually helped a lot and it's just so much faster than typing. What makes this so cool is it doesn't just dictate every word you say. It analyzes it all with AI to clean it up automatically. If you misspeak, say "um" a lot, or change your mind mid-sentence, it will fix that intuitively on its own. So, I'll just give an example. "Hey, John. Wait, who am I talking to? Sarah, could you send over the uh documents? No, no, no, not documents. Videos. Could you send over the videos?" And from that nonsense, it knew I wanted to say, "Hey, Sarah, could you send over the videos?" So, this is awesome for basic things like texting and emailing or for prompting in ChatGPT or also something like going on a walk and talking through an idea or instructions you need to send to someone or anything you do. Then you have it all cleaned up automatically. So, that one is just massive for productivity. They do have a free plan for a limited word count per week. That way, you're able to test it out. You may even be able to just stick with that depending on how much you use it. I love this one.
Next is Granola. This is an AI meeting assistant, and there's a lot of these out there. I've used a bunch of them, but the thing I like most about Granola is it doesn't have to have a bot join the meeting with you. There has been a growing annoyance around that. Some people won't even let those bots into the meeting anymore. Granola uses the system audio to create a transcript of the meeting, so no weird bot has to join. Their notepad is really cool, too. It works like Apple Notes, where you can scribble basic notes during the call. Then afterwards, it uses AI to merge the information from the transcript with your notes. It just weaves it all together in a more structured way. So that's awesome. And then the chat feature has the context of what you're working on when you ask questions like, "What did Sarah say about the budget during our meeting?" It could pull that direct quote. This one also has a free basic plan, then paid for unlimited notes and advanced integrations if you find it valuable enough.
I've been using the Comet AI browser from Perplexity for a little while, and I'm finding more and more use cases for it. I know it is a giant hassle to switch browsers. Like I still love Arc browser. I have so much organized and saved in it, but an AI-native browser does feel very futuristic and can save a ton of time. I'll go through just a few use cases to show you how this can help. I'll start for one with YouTube to open up the AI features. You come click the assistant. Then it has the obvious like summarize this video, key takeaways, and you can just chat with the information in the video. But beyond that, say there's a video where all of the chapters don't actually say the names of the AI tools. A lot of people do that. They'll say "tool one," "tool two," "tool three." They never tell you the actual name to try to get you to watch the entire thing and not skip around as easily or scan through to skip over tools you already use. So I will ask Comet, "Give me the full list of tools covered in this video with timestamps." Now down here, it has a table with the original chapters and the timestamps, then another column with each specific tool that was listed in those chapters. So that is super helpful. Personally, I always leave very precise chapters so people are able to easily skip around if they want to, but my goal is viewer satisfaction over trying to squeeze out retention.
Another use is being able to chat with websites. It's basically like having a custom chatbot on every site you go to. Comet can also research and analyze websites to help you plan or design your own or search what's working for competitors. So, I'll just demo this on our course platform. "Does the landing page clearly communicate its unique value proposition? And is the layout both visually appealing and intuitive? What could be improved?" So, I analyzed the whole site and the copy everywhere and actually has some pretty good tips on here. I could look at my email, come over to the promotions tab, and ask it to unsubscribe from all of these. So, it is going back and forth opening up emails and unsubscribing from each. I could let this run in the background while I'm doing something else and come back to a cleaner inbox. That's just a few really quick things. There's a lot more I've been finding it useful for, and each person will have their own uses.
NADN is one of the most powerful platforms for building AI agents that I'm sure most people watching have at least heard about by now. But we actually use it in the business in a bunch of different ways. One of those is for newsletters. Between me and Saj's YouTube channels and our course platform, we have just a ton of content that we do massive amounts of research for. So, it makes sense to reformat that for different platforms. For newsletters, we just drop those into a form and that kicks off a process in NAD to extract all the information and key points and convert that into a newsletter-friendly format. There is still human review after that, but it does save a ton of time. There are endless other things you can do. NAND is extremely powerful and customizable for everything from customer service and sales to content creation to logistics. We use it for most of the more agentic automations we have. It is such a big platform to cover, so it's hard to do any real tutorial in a short section like this, but I have a full NAN crash course on the channel that goes super in-depth. Learning agents and automations is one of the most valuable skills in AI right now.
And another related tool I use is Hostinger. That's how I have my NAD set up. So, there's two main ways to set up an NADN account. The easiest is the cloud version where you just sign up on NADN's site. After the free trial, it's $20 a month. But since NADN is source-available, not technically open source, but source-available, the other option is self-hosting. That gives you more options, more control, and lower costs. And there are two main ways to self-host. You can do a local install like with Docker, which is fairly technical with lots of configurations, or a virtual private server, a VPS. This is a much easier middle ground, and it's what I use myself. So, it's self-hosted, but simple to set up and cheaper than N8 Cloud. And Hostinger is by far the easiest VPS option I've seen because unlike most providers, they've built a one-click installation template for Naden. So, you skip almost all of the setup. To do this on Hostinger, go up to services, then down under VPS, select the self-hosted NAN option. Pick a plan. The KVM2 plan is more than enough for everything I've needed to do and a lot more. And they are the sponsor of today's video, so make sure to add the coupon code Futuredia. You'll get a nice discount. After that, standard billing and setup process. Then you'll see the dashboard. And since I selected the NAD option, it's right up here. To open it, click manage app, set up your NADN account. You can skip through this screen, but this one with the free license key. Click this button to send the key. Get that from your email. Go to your plan and click enter activation key and paste it in. And that's it. You are up and running with your own self-hosted version of NAD, one of the most useful and time-saving AI platforms.
Now, I'm going to change course a little bit and jump into some of the creative tools I use for content creation. And I'll start this section with more of a category than a specific platform. A big issue of working with AI video and image tools is they're constantly having updates and new releases. No tool is ever the leader for very long or at everything. So, you need to use a mix and that mix is always changing. Most of the leading tools also offer access through their API and there's platforms that aggregate all of those APIs in one place. So rather than juggling five to 10 different subscriptions and trying to remember to cancel ones you're not using anymore, instead I usually use a platform that offers them all in one place, and there are quite a few of these. Honestly, I wouldn't say I have a favorite, but I like Freepick a lot. I tend to come back to them. Creta and Leonardo are great, too, and have a lot of their own unique features and tools added in, which is nice, although I haven't used either for a while, so I'm not quite up to date. But another that I've started using more recently is Art List, which used to be just a stock platform for stock images, video, music, templates, LUTs, and things like that. It's been one of the top stock platforms for a long time, and now they've added in the AI image and video models. So, I like that because it also replaces another subscription I had since I use stock media all the time. Now, I've had sponsorships with some of these platforms at different times, but really the different platforms are good for different people. Free pick for easy access to all the different models, really clean and easy. Creta or Leonardo if you want to add in additional unique creative tools or Art List if you also use stock media. There's a lot of other tools that I don't have as much experience with. Whichever one you choose, just having one platform to go to for every model is really nice.
Another platform like this is Replicate. Not the same nice UI as the previous platforms have, but it has a ton of different tools across various use cases, including all the image and video tools I just mentioned. On top of that, they also have LLMs, audio, music, vision, fine-tuning, and a lot of smaller specialized open-source tools. It's like a library of AI tools with thousands of different models you can easily search for and use. It's connected to GitHub. You just load your account with some money and pay by usage rather than a subscription. An example of a cool one I've been using on Replicate is MM Audio. How this works is you can drop in a video that doesn't have sound, click run, then it will generate matching sound effects for you. THAT'S GREAT. ALL THE SOUND EFFECTS match pretty well. Not perfectly synced, but definitely has the right feel. Some of the video platforms generate the audio and video simultaneously now, but not all of them yet. So, this is a quick way to get some audio. It's not perfect complex sound design or anything, but for a quick little video that you're posting on social media and don't want to spend a ton of time on, or for like quick B-roll, it can be great and add some depth.
And I recently came across a new platform called Mirell. It does the same thing, but in my first test, it has actually been much better. I haven't been using it for long enough or had like a big project to run through it yet, but it has been great so far. With that same idea, you choose a video. I will choose automatic. Then it analyzes the video and generates sound effects and music on separate tracks. That's really nice. Then you can control the volume of each track separately. It also generates multiple options for each track. So, I'll start off with the sound effects. So, those all sound pretty great. And then I'm going to turn that off and test out the music. It sounds good. Just some kind of ambient background music. I'll turn them both up. There we have it. That looks great. That adds a lot more depth to this. So, I'm definitely going to be using this one a lot more.
Another tool I use a lot is Magnific. With AI image tools, it's very common to get low-resolution images or just images with some errors or warping. Magnific has a suite of upscaling tools that solve each of those issues. So, they offer precision upscaling for staying close to the original like a traditional upscaler, but those have been around for a long time. The more unique side is creative upscaling, which upscales while reimagining the details. It can add an insane level of detail this way. And there are custom controls to guide the generation to be more or less like the original. It's an extremely useful tool if you regularly work with AI images. It enhances the quality massively. Now, it is on the expensive side, but it is so far ahead of the competition that it's the one I use.
In the same realm is Astra by Topaz. This is a creative upscaler for video. With that same idea, AI video often has low fidelity or warping issues. Astra upscales that to clean everything up, and it is pretty insane how well it works a lot of the time. And they do also offer the precision upscaler for standard upscaling. This one is also on the expensive side, but extremely valuable if you work with AI video a lot.
And while I'm talking about creative tools, I have to cover Nano Banana from Google. That has become a staple in my workflows for thumbnails, consistent characters and scenes in AI videos and custom graphics. You can upload an image and ask for any change you want, and it delivers amazing results. And they are supposedly dropping Nano Banana 2 any day now. So hopefully it solves any of the little gaps in the first version with this update. I've done multiple deep dives comparing it to all the other AI image editing tools across tons of use cases. It was the winner in most of those already, not all of them, but maybe it will be the leader in everyone with this new update. Either way, it's already an amazing tool, and I regularly combine it with Magnific.
And while we're on the topic of Google, there's a few other tools I use from them all the time. Notebook LM I use almost every day for organizing my learning and interacting with it in different formats to help retain the information. This is the most underrated AI tool in my opinion. You upload sources to a notebook in many formats. You can use PDFs, text files, YouTube videos, and websites all in the same notebook. If you want it to expand to more sources, you can click "Discover Sources" and add related material. Then when you chat, it uses all of that as context before answering and cites exactly where the info came from so you can double-check it. That helps with one of the biggest problems with most chat models: hallucinations. They don't disappear completely, but they drop way down when the model's grounded in your own data. And from there, you can do some really cool things. You can ask it to turn your research into a quiz, a video overview, a report, a mind map, a set of flashcards, or the podcast mode, which went super viral when it came out. It creates a natural conversation between two AI hosts based on your notes. And you can even call into that podcast and ask questions in real time.
>> Today, we're uh really digging into retrieval augmented general.
>> Oh, wait. Someone wants to join. Hey, go for it.
>> Can you explain how vector databases work in simple terms?
>> Oh, that's a great question.
>> That's the core component of Rag.
>> And we absolutely can break that down simply.
>> Okay, let's unpack this.
>> Think of a vector database less like a traditional database.
>> You know, with rows and columns.
>> Exactly. Instead, think of it as a really specialized address book for ideas.
>> An address book for meaning.
>> Precisely. It is the best tool I've found for actually retaining information because you can see it, hear it, and interact with it all in one place. I'm just a big fan of this one. I even did a full video just on Notebook LM and every feature it has recently if you want to go deeper with it.
Something else I've been doing in every LLM is creating little mini apps and automations. This works similarly in Gemini, ChatGPT, and Claude, but I'll show it in Gemini. It's extremely simple to build a small automated workflow that utilizes multimodal tools. You can streamline processes for all types of work, usually with just a few prompts. And this is on the free plan. I just select the canvas feature and typed a couple sentences describing what I wanted: a title exploder combined with a thumbnail generator. And just from that, it generated the entire mini app. Now I drop in an idea. It gives me 20 potential engaging titles. Then I select one. Then I can upload my face and a style reference. Then it will generate me a thumbnail based off of those. After that, I can edit it further using natural language. And this took only a few minutes to build. And you can do the same type of thing in ChatGPT projects. I actually covered that in my projects video. Or also with Claude skills and artifacts. Doesn't matter which LLM you use. They each have some sort of option like this where you can combine a knowledge base and custom instructions to create an automated workflow or create a mini app to automate something. I think basically everyone has a way they could use these to save time.
If you want to build something more advanced than this, Google also launched a new vibe coding feature inside of AI Studio, which is a platform I use a lot. It's a whole suite of Google AI tools like Gemini, Nano Banana, and more, all for free, but with more customization and the ability to lower the safety restrictions. That new vibe coding tool is under the "Build" tab. You can build full apps and games using natural language. Then there's also a visual option that lets you draw and add annotation boxes and text to ask for changes. Sometimes that's an easier way to get what you want than trying to explain it in words. So if you want to build something beyond what you're able to build in that canvas feature I showed, jump in here and it offers much more capabilities and customization.
And if you want to go much deeper into learning AI, we do have a full course platform on Futuripedia with over a thousand lessons across over 30 AI courses. You'll find full learning paths on everything from ChatGPT to video generation to coding with AI and everything in between. We even have full courses on many of the tools I talked about in this video. It's all included in one subscription. You can get a 7-day free trial using the link in the description. Or if you want to dive deeper into some of the Google tools I covered near the end, check out this video with 33 free Google AI features. There's a ton of valuable stuff in there.