Create Stunning Logos with AI (for Free)

Use Stable Cascade, a free AI model, to generate logos without paying for Midjourney. Run it locally on your GPU or use Hugging Face Spaces online. Convert the raster output to vectors with Vectorizer.io, then polish in Illustrator.
Questions this video answers
February 12, 2024
February 15, 2024
$10.80 per hour
$5.40
$5 for a single week or $10 per month
Chapters
Want to know how to make really good AI logos using completely free AI tools? Some of these AI art tools have gotten really, really good at making logos for you. Logos like this pirate logo, or this wolf logo, or this old school hockey logo, this Mike's Repair Shop logo, this reimagined Montreal Canadiens vintage logo, reimagined San Diego Padres vintage logo. A cool logo of some red lips, an astronaut riding a Hippo logo, because why not? Or this sort of ABST MW logo, or this logo for Niko's plumbing that uses absolutely no existing trademarked ip. I actually tried reimagining the Future Tools logo and a couple of them weren't half bad. Now about a year ago I put out this video on my YouTube channel creating AI generated logos using Mid Journey. It is the second most popular video on my channel and since the time that I released that video, Mid Journey started charging to use it. And we've gotten a ton of new AI art generators and one brand new one that just came out the week that I recording this video is actually really exceptional at creating logos. Hat tip to my buddy Angry Penguin here, AKA AP over on X, who showed off that it was really good at making logos by making his own AP logo. And here's another variation he made. All of these logos you've seen so far were created using a completely free open source tool. The tool that I'm referring to is Stable Cascade put out by Stability AI. You can see that this art generation model came out on February 12th and that's 2024, just in case you're watching this a year or two down the line. And just overall it's a really good AI image generation model. It doesn't do realism very well. I would say that Midjourney V6 and Imagen from Google are both much better at realistic images, but this is still a really, really solid model. And as you can see, it does text well. And from the examples I just showed you, it also does logos really, really well. The Stability website has a blog post that goes into all of the technical details of how this works, but what we really care about here are these graphs. So the prompt alignment, that is how well it adheres to the prompt that you give. It beats out Playground V2SDXL, Turbo SDXL and Worst Gen V2, which by the way, that Worst Gen architecture is actually what this model was sort of built on top of. So this version of it is quite a bit better, I suppose, than the previous version of it. As far as aesthetic quality, it ranked a little bit lower than playground v2 but much better than sdxl turbo sdxl and the worst gen v2 as far as speed when prompting it is way faster than SDXL and Playground V2, but not quite as fast as SDXL Turbo. It also looks like it'll work with control nets in painting and even variations given an original input image. Again, I'll make sure this article is linked up below, so if you do want to get into the weeds with it, you can read more about it, but you probably clicked on this video because you want to know how to get logos out of this I'm going to show you two ways to be able to use stable cascade 100% for free. Now the first way is by using the browser called Pinocchio. You can download this over at Pinocchio Computer. It's basically a platform slash browser that makes it really easy to install various models on your local computer. And as we can see right here, Stable Cascade is verified and working using this Pinocchio platform. If you're not at all familiar with Pinocchio, make sure you watch this video here called AI Swiss Army Knife Free Tool does everything. That video is all about this Pinocchio platform. It lets you install pretty much any AI model on your local computer, assuming your computer is capable of powering that model. You can also see that this Pinocchio app works on Windows M1, M2, M3 Max, Intel Max, and even Linux, so it should work on whatever computer you want to use it on. Now I already have Pinocchio installed, so I'm just going to go ahead and search it out on my computer here and open it up. You can see all of the various AI models I've already got installed inside of Pinocchio here. And again, if you need a step by step tutorial on how to install Pinocchio, I do walk you through it in this video right here. The AI Swiss Army Knife video. With that being said, it's pretty self explanatory. Now opening up Pinocchio. If I click on this Discover up in the top right here you can see all of the models that are available at the time of this recording. Stable Cascade is towards the top because it was just recently released, so I'm going to click on that. It gives me the option to download Stable Cascade. So I'm going to go ahead and download it. After a few seconds it shows me Stable Cascade is downloaded and now I can install. I'll go ahead and click Install and then I'll click Run and now it's going to start the install process. This might take a few minutes because it is going to download some things from the web, including some various image weights and things like that, which are multiple gigabytes. So step away, grab a sip of water or something and just let it do its thing. After a few minutes, when it's done installing, on my homepage of Pinocchio here, you can see I've got Stable Cascade. Now I can click Start, and then up in the top left, simply click Run. And for whatever reason, it doesn't launch automatically. For me, I think this is probably something that this Pinocchio app will fix. Eventually you would think you'd press Run, it would run some code and then launch the browser, but for whatever reason, it's not doing that. Instead, it's giving us this thing that says run on local URL and then you can see the URL here. If I click on this URL, it opens Stable Cascade in my browser here. Now, this is totally local. This is not running in the cloud. This is not online. I could unplug my Internet right now and this would work. And if I give it a prompt like the suggested example, the spirit of Tamagotchi wandering in the city of Los Angeles, click run. And this is the final image we get. Now, while actually recording my videos with my tool called Vmix and trying to run this at the same time, it actually really bogged down my computer. You can see my GPU usage was at 91%, while Vmix, my recording software, was 7%. My GPU was pretty much maxed out. Here's another screenshot showing the process running on my computer. With my GPU totally at the very, very top of the line there. Even with my Nvidia Gefor 4080, with 16 gigabytes of RAM, it took about 20 minutes to process this. So Pinocchio is a good way to install it on your computer for free. Just keep in mind you need a pretty powerful GPU to be able to run this properly. It's also quite possible that inside of Pinocchio, the way the install process happened, maybe it wasn't totally optimized. I'm not sure. It just took a really, really long time. So for most people, you're probably going to want to go the other route that I'm about to show you, which is using Hugging face. If you go to huggingface, search for Stable Cascade. You'll see up here, Stability AI Stable Cascade. We can click in. See this model. The information that's here is pretty much the same information that's on their GitHub page. But then over on the right you can see a little area that says Spaces Using Stable Cascade and there's a handful of spaces using it right now let's go ahead and use the multimodal art Stable Cascade. And you can see we get pretty much the identical user interface. We're just in the cloud now. We're on Hugging Faces Clouds now. Every time you enter a prompt on one of Hugging Faces Clouds, at least with Stable Cascade it's going to look for a GPU to process it on. Sometimes you do have to wait a while for a GPU to become available before it starts processing. So if you have a really good gpu, install it locally. You don't have to wait for a GPU to open up. Otherwise, use the Hugging face version. If you want it to run as fast as possible. You can come up to these three dots up here in the top right corner of Hugging Face and come down to duplicate this space. Under space hardware, select one of the GPU options down here. But keep in mind that it does cost you some money to use. So let's say you grab one that's $10.80 per hour, their most powerful GPU available, and you end up using it for 30 minutes. You will end up paying about $5.40 to be able to use that GPU. But let's go ahead and just try it on the free version, see how it's going right now. And again, this video is all about creating logos with it. And you don't have to do really any prompt engineering anymore. If you watch this old logo video that I made here on YouTube, there was quite a bit of prompt engineering involved. You had to enter a whole bunch of special extra characters to get the image just right now it's pretty easy. I can just type a logo for an auto repair company called Road King. I'll click run. You can see up here we get our waiting for GPU but pretty quickly we got successfully acquired. So the GPUs right now as of recording this are coming fast. And you can see as I'm talking, this is generating in real time. I didn't speed this up. That's how quickly it moves on the hugging face space. Got a couple of the letters wrong here, added an extra O and an extra N. But not a bad looking logo for just a simple prompt here. I could run it again, see if it gives us a better one second time around. Check that out. Got the words perfectly right and It's a decent looking logo without much extra effort on my part here. There are some advanced options down here. You can add a negative prompt, you can select a seed. I like to just leave it on random seed. Mess with the width, the height, have it generate more images. You can set a guidance scale, inference steps, and a few other options that you can play with and see how it affects your output. Let's do a logo with a wolf and the word wolf with an e on it. Go ahead and run that first time around. The wolf looks great, but the words not perfect. Let's go ahead and run it again. In my opinion, an even better wolf image still got the letters wrong, but I mean that's easy enough to pull into Photoshop or something like that and change the letters. Let's go ahead and run it one more time. I'm confident it can figure out how to spell wolf with an e at the end. And third time's a charm. We've got wolf spelled the way I wanted it to. It made the f a lowercase, but it still got there. Let's do a logo for a coffee shop called Yummies. Look at that. That's pretty dang solid. I'd say that looks more like ice cream. I guess you could argue it's maybe whipped on it. Pretty dang good though. And the beautiful part about AI, especially free to use AI, just keep hitting the generate button, keep hitting that run button until you get the exact image that you want. Now, another thing that's really good about Stable Cascade is the prompt adherence. You could give it additional details and it's pretty good at following them. Let's test this longer prompt. A logo for a coffee shop called Yummies with a blue mug and steam rising from it. The word Yummies can be seen in the steam. Let's see if it can adhere to that prompt here. It got most of it. It got the blue mug, it got the steam rising out. It put yummies on the mug instead of in the steam, but pretty close. Let's just run it again, see what happens. Second time, same thing. I've tried several times now. No matter what I do, I can't get the steam to spell the word yummies. But it does follow the blue mug with steam rising from the mug every single time. But with enough testing different prompts and hitting the run button over and over again, I'm sure you could get it if that's your ideal style. Let's try a logo with a monster and the words scare, floor not bad. It didn't nail the spelling on the word floor the first time around, but let's hit it again now. Look at that one. That one pretty much nailed it. It's just really amazing how quickly we got to a point where prompt engineering, at least with AI art, has become less and less of a factor. We just tell it we're looking for a logo. Here's what I want in the logo. Here's the words I want in the logo. It'll make you something. If you're curious about some of the prompts that were used in some of those logos I showed you in the intro. Well, a lot of these were made by John, my creative director on my team. And here's the prompts that were used. So this here is a minimalistic 2D simple professional eye logo for this hockey mask image here it was simply modern simple sports logo of old school hockey. For Mike's Repair Shop we have professional red neon logo. Mike's Repair Shop did add an extra repair word in there, but not bad for this one right here. John's prompt was Montreal Canadiens vintage logo rework. He then took on a Padres vintage look with the prompt San Diego Padres vintage logo rework and came up with this one for these lips right here. Logo of red sexy lips. And this pirate logo that I thought was really, really cool looking the whole prompt logo of pirate. And it came up with this image. Not bad. Now, if you want an extra step and you really, really want to get polished with your logos, there's a couple other sites where you can actually convert these into a vector graphics. There's Vectorizer IO. This site costs about 5 bucks to use for a single week or 10 bucks to use per month. Vectorizer AI, 10 bucks a month to use as well. Any image that you upload into this tool or this tool, Vectorizer IO or Vectorizer AI. Well, convert the image that you just created into a vector graphic, which makes the graphic infinitely scalable without quality loss. And you can pull it into tools like Adobe Illustrator and clean it up and dial it in even more. So if you are a professional graphics designer, you do this for a living, you can actually get your base built with something like Stable Cascade totally for free. Pull it into Vectorizer IO or Vectorizer AI, convert it to a vector and then finish it off, polish it off and really dial it in using something like Adobe Illustrator. There's how you can make some really cool AI generated logos using Stable Cascade completely for free, either locally installed or through the hugging face space. That's currently available. After this video goes live, there might be a little bit more traffic on that hugging face space, so it might be harder to get GPUs. If that's the case, you can always duplicate the space. Pay for one of the GPUs at like two bucks an hour. If you only use it for 20 minutes, you end up spending, you know, 50 cents or something like that. So hopefully you found this helpful. Hopefully you enjoyed it. If you want more tutorials like this one, you want to stay in the loop with the latest AI news, latest AI research, and just nerd out about really cool emerging tech as well as AI. Make sure you like this video and subscribe to this channel and I will make sure more videos like this show up in your YouTube feed. And of course, if you haven't already, check out Feature Tools IO. This is where I curate all of the coolest AI tools that I come across, update the AI News page on a daily basis, and I have a completely free newsletter. You join the free newsletter, you get access to the AI Income database, which is a ongoing database of cool, interesting ways to make money using AI tools. Plus you get the weekly newsletter with just the coolest tools. I come across the most interesting AI news and it's all totally free. Just head over to FutureTools IE and I'll hook you up there. Thank you so much for tuning in. I really, really appreciate you. I have so much fun nerding out, testing out these tools and sharing it with you. Hopefully you enjoyed this one and I'll see you in the next one. Bye Bye.










