Transcription
Hey guys, this is my Android phone and it's running fully uncensored AI completely offline. No internet, no cloud, and no complicated setup. It's just a normal app like anything you install from the Play Store. But it's running a real AI model right on the device. Watch this.
I'm turning off Wi-Fi and mobile data and yeah, the airplane mode is on. And now I'm going to ask it something that ChatGPT would never answer. And look at that. It just responds directly. No warnings, no lectures, just a straight answer. This is a real uncensored AI running locally on a phone and everything you see here works exactly the same on iPhone or iPad, too.
A few days ago I posted a video showing how to run an uncensored AI from a USB on Linux, Windows, and Mac and it blew up. I got a lot of comments asking for Android and iOS. So I built this. I turned that whole setup into a simple app. Honestly, it took way more time than I expected, but I really wanted to make it easy for everyone. And the best part, it's fully open source, so anyone can explore it, modify it, and even build on top of it. And if you appreciate the effort that went into this, a quick subscribe would really mean a lot.
All right. Open the GitHub link in the description and download the APK. If you're on iPhone, the iOS setup is also linked there. Download the ARM 64-8A build. This one works on most Android phones. Once it's downloaded, install it like a normal app. Once the app is installed, just open it up and you'll see the main chat screen. But we can't chat yet because we don't have any model downloaded. So tap on the models tab at the bottom. This is where you'll see all the available models. I recommend starting with Gemma 2 2B because it's the smallest at around 1.6 GB and it runs well even on phones with just 4 GB of RAM. I've already got it downloaded, so let me just import it.
So tap on import, then choose import GGUF file. Here's the Gemma 2 model around 1.7 GB. I'll select it. Now wait a few seconds while it copies into the app folder. And yes, the import's done. You'll now see the load model pop up. Right now, the model is just sitting in your phone's storage, but the moment you tap load, it gets pushed into RAM. And that's why more RAM means smoother performance and faster responses. And yeah, the model is now loaded into RAM. We're ready to start chatting. You can see it also detects which model is currently loaded. Let's test it. For the first message, it might take around 10 to 20 seconds, so just wait a bit. And there it is. It's responding. Now I'll turn off Wi-Fi and mobile data just to confirm everything works offline. And let me ask again. So yeah, it's replying even with no internet at all.
And honestly, this whole idea of having something fully private and isolated actually fits perfectly with today's sponsor, Plugmate, who's doing something very similar. While this app gives you private AI on your phone, Plugmate takes it even further. It's basically a tiny device that turns any phone, laptop, or even a TV into a completely separate Android system. Not an app, not an emulator, but a real isolated Android running on its own hardware. So your main device becomes just a screen and everything else like apps, data, storage stays inside this device, which means better privacy, better control, and zero mixing with your personal data. I actually made a full dedicated video on this where I tested everything in detail, setup, features, and real usage. So if you want to check that out or if you're thinking of getting one, I've put all the links in the description. Now let's get back to the setup.
All right. It's time to test if this is actually uncensored. I'm going to ask the same question to ChatGPT first. And look at this response. It just dances around the question, gives me warnings, tells me why it can't help, and then gives some watered-down generic answer. Now I'll ask the exact same question to our local AI on the phone. And there you go. It just gives a direct, raw, honest answer. No moralizing, no refusals, no corporate safety disclaimers. Just a straight response to what I asked.
From the sidebar, you can access your chat history, delete the older chats, or start a new chat anytime. From model section, you can download any model you want. Just make sure it matches your phone's specs. And from the settings, you can tweak the temperature to control responses or change the system prompt. Everything is super easy to use. There's also a local API server option. If you're a developer, you can connect other apps to this model using OpenAI style endpoints, which is honestly really powerful. But my favorite feature is this, importing models using a URL. If the model you want isn't in the list, no problem. Just grab its GGUF link from Hugging Face, paste it here, give it a name, and add it. And just like that, it shows up like any other model. Now you can download it, load it, and start chatting with it. I even tried the latest Gemma 4 and it runs surprisingly well. Of course, it's a bigger model, so it's a bit slower compared to Gemma 2, but that's expected because of the higher number of parameters. So yeah, this is basically a full offline AI setup right in your pocket.
On iOS, you can't install it like Android. If you're a Flutter developer, you'll already know how to run it. If not, just follow the GitHub guide. I've kept it simple and step-by-step for iPhone and iPad. So that's it. You now have a fully uncensored AI running locally on your phone. No internet, no cloud, and no content filters. Everything stays on your device completely private.
If you found this video useful, hit the like button and subscribe because I'm going to keep building more tools like this. And if you haven't seen my USB AI video yet, go check that out, too. I'll link it in the cards and the description. Thanks for watching. This is Jarvis signing off.