Pinocchio
What if I told you there was a single, simple platform to run things like automatic 11-11 for stable diffusion locally on your computer or animate diff to create really cool AI animations or tools to make voice clones or face swaps or run large language models that are open source directly on your computer and it works on Mac, it works on Windows, and it works on Linux, and it’s completely free to use?
Well, that exists, and it’s called Pinocchio. You can find it over at Pinocchio.
This is a tool that I’ve had my eye on for several months.
Many people have told me about it, I’ve taken a peek at it, but I’ve been hesitant to play around with it for a couple of reasons.
First of all, this Pinocchio website here is a really basic website.
It’s got a headline. It’s got a YouTube video. It shows you what kind of tools you can install, and that’s the entire website.
It’s created by this person who seems to be either anonymous or pseudo-anonymous, who goes by the name “cocktail peanut.”
The fact that the site was so simple, and I didn’t really know who was actually behind this platform, has kind of put me off from trying it out.
However, this has been around for several months now, and a handful of people that I really trust, who actually understand the code underneath it, have looked through it, vetted it, and told me that this is a legitimate tool that works really well.
Cool Features of Pinocchio Tools
Once you’ve got it set up, you can do things like face fusion to swap your face with anybody else’s face inside a video.
You can do things like illusion diffusion, which was a trendy style of image a few weeks ago where you got these swirls that blended with house pictures or these checkerboard patterns that actually had a little image of a town inside of it.
It’s got a voice cloning tool. I don’t know if it’s pronounced “Ki” or “Ki,” but it’s a really good tool to clone really any voice, similar to what you could do with 11 Labs.
It’s got token flow in it, which does things similar to what you get out of Gen One from Runway.
It’s got model scope video to video and image to video if you’re running with an Nvidia GPU, so probably not going to work on a Mac.
It also has “comfy UI,” which is a tool to generate AI-generated images with stable diffusion.
I’ve kind of steered clear of it because these spaghetti bowl-looking images have really overwhelmed me, but I think I’m this close to diving in a little bit deeper on it and playing with it some more.
It’s got animate diff, which allows you to generate AI-generated video, and of course, you can even do things like automatic 11-11.
Which is the interface that you normally see with something like stable diffusion, and this is just a handful of featured scripts that are recommended. There’s a lot more scripts that are available as well.
Installing Pinocchio tools
Now that I’ve had some friends properly vet it for me and tell me that it’s a safe tool to use, let’s go ahead and install it.
Download it and try a few of these add-ons and see if it’s as good as everybody claims it is. So, I’m going to go ahead and download this. You can see that we do have options for Windows M1 and M2 Max Intel Max and Linux, so let’s go ahead and download it for Windows.
Here I have a folder on my computer called AI. I just created a new folder called Pinocchio, so I’m going to go ahead and paste it in here and unzip it to this specific folder.
Now it does say Windows protected your PC, but I’m going to go ahead and run anyway to make sure it’s installed in this new folder that I just created. I’m a dark theme kind of guy, so I’m going to go ahead and set it to Dark theme and click save.
Now the first issue I’m running into is that it seems to only want to let me install it in this folder. I can’t install it on a separate hard drive. So I guess I’m just going to let it install it where it wants to because whenever I change this, it throws up an error.
This looks like this is the Pinocchio browser here, so I’m going to go ahead and full-screen this. Let’s click on the Discover page.
Now it takes us back to this page where we have all of these various tools that we can install directly into Pinocchio. The first one that I think Stable Diffusion Automatic1111 makes sense to install is probably automatic 1111. This is the one that I use the most to generate AI images, so we’ll go ahead and click this and select download on stable diffusion. It’s telling me there’s some other stuff required to install, and I believe if I just click install, it will go ahead and install all that for me.
Installing Pinocchio tools on my computer
Now I do have all these things installed on my computer, but it must need to install them specifically for this browser. So I’ll go ahead and click install, and you can see down here it says installing cond. So it looks like it will go through and install each of the required tools.
All right, so it looks like it finally installed all of the prerequisites. It took a good 7 minutes or so to do that, and then it put me back to this page. So, I think I need to click download again. We’ll go ahead and download the SD web UI. Pinocchio dog it. Now, it looks like it’s asking me if I want to install stab defusion XEL or stable diffusion 1.5. I kind of want them both, but let’s go ahead and start with Excel.
So, it finished downloading the stable diffusion models. It took about 3 minutes to download the models. Now, it looks like it’s installing all of the necessary files to actually run stable diffusion. So that took about six more minutes, and then it just launched stable diffusion right inside of my main browser. So it didn’t seem to launch stable diffusion inside of this Pinocchio browser here. The Pinocchio browser just seems to run the install for you. Once it’s installed, you get a little launch button here, and when you actually do launch it, it opens it in whatever your default browser is.
Under my stable diffusion checkpoint, you can see it added the stable diffusion XL base model, and I should be able to let’s set the sampling steps to 40 here and do a prompt: “a wolf walking in a snowy Forest.” I click generate, and there we go, we got a seemingly Five-Legged wolf walking in a forest.
But stable diffusion isn’t the only thing that Pinocchio makes it really easy to install. In fact, there’s a lot of other things that I’ve come across on GitHub that I’ve wanted to play with but haven’t taken the time to learn how to set them up on my computer. A lot of these tools are just built into Pinocchio where you can install them really easily. Tools like Face Fusion, which is a face swapper tool, and this xtts or Ki or Ki (I don’t know how it’s pronounced) where you can clone a voice, similar to 11 Labs.
There are other tools like token flow and animate diff and comfy UI, which are all tools I want to play with over time. But let’s go ahead and set up Face Fusion where we can do our own face swapping. I’ll click download on this one because I’m running on an Nvidia GPU. I’m going to select install for Cuda and then click install. There are a couple of prerequisites here that it’s going to go ahead and install for me in this process.
Successfully Installed Pinocchio tools on my computer
All right, install success. That took a good, I don’t know, six or seven minutes here. Okay, so I’m going to set the face swapper and the face enhancer. I’m going to leave these as the default. Make sure my execution provider is Cuda since I am using an Nvidia GPU. If you’re not, I’d set it on CPU execution. Thread count, I’m going to put it at 16 here, Q count One max memory. I’m going to leave everything kind of the default.
Now the source, this is where I’m going to drag and drop my image. I’m going to make myself look like Dave Grohl.
Here as my test, so I’ll go ahead and put Dave Grohl in as the source. For the Target, this is where I’m going to toss my video. For video, I just have this random clip of myself doing an intro to one of my other videos. So I’m going to go ahead and pull that in as my target here.
I have face analysis left to right. Face analysis age is adult, gender is male. I’m just going to prop off a bunch on the front here. This is funny; I can actually sort of skim through my preview frames here and see what it thinks my face would look like if I put Dave Grohl over it. The most notable feature here being the teeth. I think that’s pretty hilarious. But let’s go ahead and click Start and see how it renders out this video with Dave Grohl’s face.
All right, so it took roughly 7 minutes, and here’s what we got from our face swap. I have a habit of collecting AI research and just stockpiling really cool stuff that I come across. It really thinks Dave Grohl’s teeth are pretty pronounced. I like it though.
Let’s set up Illusion Diffusion.
All right, so we got Illusion Diffusion up and running. It seemed to work on the first try. This one was actually created by a buddy of mine who goes by AP or Angry Penguin on Twitter. So if you’re not following him, make sure you are. He shares a lot of cool AI stuff on his Twitter.
But this one, you’ve probably seen a whole bunch of stuff go viral. Let’s go ahead and try a checkerboard pattern.
Here as our input illusion, and for the prompt, let’s do an urban Downtown City. We’ll leave the negative prompt as low quality, and I’ll leave the advanced options as the defaults. Let’s just go ahead and run it, and it was almost instant. However, the illusion is a little more subtle than I’d like. You can tell it did sort of follow the squares, but I think I want to crank the illusion strength up a little bit. Let’s just set it at one and see what we get out of it, and this one’s almost immediate too. Like these images generate really quickly. You can see it worked again; you kind of got to squint at it to really see the checkerboard, but it’s definitely working. Let’s try it with the swirls, just for fun. Run that one. The illusion is definitely more apparent with the swirls, but also, I don’t know how common clouds like this are.
We’ve got our screen to play around with token flow. So I have this b-roll of a robot typing at a computer. See if I can make it look like a zombie typing at a computer. I’ll drag this video into my input video here. Let’s do a bloody, decaying zombie typing at a computer since it is Halloween on the day that I’m recording this and click edit your video.
Zombie robot
So it took about a minute and a half here, and it generated a 2-second clip. Let’s see what it gave us. Definitely much more zombie-like. It looks sort of like a zombie robot. It also kind of messed with the aspect ratio a little bit, but there are some other options we can play around with here under the general options and plug-and-play parameters. More stuff that you can play around with here, but this gives you an output similar to what you’d expect from a Runway gen 1.
Open Orca
All right, so it looks like it is done downloading. I’ll go ahead and click on this refresh button up here, and now we have open Orca mistal 7B as an option. After loading the model, I had to hit reload, and now it says successfully loaded open Orca. Let’s jump over to our chat window here and say, “Write me a poem about wolves who love AI.”
Just like what you’d normally be used to with one of your other chat apps, we get a poem about AI. However, it did cut off right here. So if I jump into my parameters tab, I should be able to up the number of tokens here and click regenerate. This time it seemed to have generated a full poem. So now we have our own large language model running locally on our computer, not sending data to OpenAI, Microsoft, Google, or anybody. This is completely local, and I can run this even without an internet connection right now.
So my final thoughts here on this Pinocchio tool:
It seems to work with most apps really, really well. However, it is sort of at the mercy of the various files that are available on the GitHub repositories for these various tools. So if, for example, xtts isn’t working with the code on GitHub, well then it’s not going to work very well on Pinocchio. They are kind of at the mercy of some of the tool creators. However, it seems like most of the tools I’ve tried work.
I was able to do token flow, which gave me a Gen 1 effect. I was able to run stable diffusion. I was able to run uaba (I don’t know if that’s how to pronounce it or not, but it’s fun to say). I was able to run illusion diffusion. I was able to run face Fusion finally. Right as I was about to stop this recording, I got another message from Cocktail Peanut here inside of Discord, saying, “By the way, I just tried bark myself, and it turns out it indeed was broken today because of the new gradio 4.0 that just went live today, which seems to have made some breaking changes.”
I rolled back the Gradio version to an older version, and now I can confirm it’s working. So by the time you watch this video, most likely the bark voice cloning is also going to be working. This is essentially a beta product. It is very, very early. They are making updates every single day to this and adding new features and adding new models that you can play with because they’re not the ones actually creating a lot of these models. Some of them are going to break sometimes. That’s just the way it’s going to be.
However, it is the easiest way to install some of these models that I’ve come across at all. You just install Pinocchio, and then any model you want to run is like another one-click install. Yes, the installs take anywhere from 5 to 10 minutes to install, but that’s usually because it’s downloading a bunch of stuff from the internet and then running an install from them.
Pinocchio Discord
I’ve also found the Pinocchio Discord to be super helpful. Cocktail Peanut himself is in there answering questions, and when bugs pop up, he seems to be looking into them and trying to fix them really, really quickly. So they are really on top of it. Man, this is cool. This is how I’m going to use face fusion and illusion diffusion and any sort of locally trained large language models. This is the way to go, honestly.
Check them out. You can find them again over at Pinocchio (p-i-n-o-k-i-o). This isn’t a sponsored video for them. This tool doesn’t even cost money. It’s free. The people building it seem to be just building it because they really enjoy building it and they wanted to make something useful for the community. There is nothing for sale here. It is just a free and really, really easy-to-use install platform that works on any computer type.
If you really want to stay in the loop with all the latest cool AI tools that are coming out, make sure you check out Futur Tools. This is where I curate all the latest cool AI tools that I come across. I also keep up to date on the AI news on a daily basis here, and I have a free newsletter where I will send you the latest AI tools, the latest AI news directly to your email inbox. You can find it all over at Futur Tools.
