No WiFi No Data No Payment, This AI Still Works
No WiFi. No data. No payment plus 100% privacy That sounds like a scam, right? Because for years, we’ve been told that AI only works if you’re connected, that every question you ask gets sent to some powerful server somewhere in the world. Your inbox is overflowing, and you need to draft a professional response, but you're offline. Normally, this is the moment you'd accept defeat and open the in-flight magazine. Not anymore.
Google has quietly released a free app called AI Edge Gallery that lets you download full AI models directly onto your phone and run them completely offline. This isn't a stripped-down, watered-down assistant it's the real deal, running powerful generative AI models locally on your device's hardware. The app is now officially available on both the Google Play Store and the Apple App Store, making it accessible to anyone with a compatible device.
The most powerful part of this entire setup is a single fact: once the model is on your phone, it works anywhere. No Wi‑Fi, no mobile data, no roaming charges, and no privacy concerns about your data being sent to a remote server. All processing happens directly on your device. It's not science fiction; it's something you can try right now.
What Exactly Is Google AI Edge Gallery?
Google AI Edge Gallery is an experimental app designed to showcase on-device machine learning and generative AI use cases. In plain English, it's a sandbox where you can download AI models mostly from Google's open‑source Gemma family and run them entirely offline. Unlike cloud‑based AI assistants like Gemini or ChatGPT, which send your prompts to a server for processing, AI Edge Gallery downloads the "brain" of the AI directly to your phone.
This means you gain a level of privacy and reliability that cloud‑based AI simply cannot offer. Your prompts, your images, and your sensitive data never leave your physical device. For anyone who has ever hesitated to ask an AI a private question about their health, finances, or business, this is a game‑changer.
The app comes packed with several core features, each powered by downloadable models that run locally. You'll find AI Chat for multi‑turn conversations, Ask Image for visual questions, Prompt Lab for text generation and summarization, Audio Scribe for offline transcription, and even a playful offline mini‑game called Tiny Garden that uses natural language commands.
What You'll Need Before You Start
Before diving in, there are a few practical requirements to keep in mind. First, storage space: the AI models you'll be downloading are substantial. The recommended Gemma 4 E2B model weighs in at approximately 2.58 GB, so you'll want to ensure your phone has enough free space before you begin. Smaller models are available if you're tight on space, but the larger models deliver more capable responses.
Second, device compatibility: the app requires devices running at least Android 12 or iOS 17, and performance will vary depending on your phone's hardware. Newer devices with more powerful processors and dedicated AI chips will run models noticeably faster. The app also requires at least 4GB of RAM to function smoothly, and devices with NPUs (Neural Processing Units) or GPUs will see significantly faster inference speeds.
Third, a Hugging Face account (if downloading from GitHub): while the Play Store and App Store versions streamline the process, users downloading the APK from GitHub will need a free Hugging Face account to access and download the models, and you'll need to accept the license agreements for each model you wish to use. This is a one‑time setup that takes just a few minutes.
Step-by-Step: How to Install AI Edge Gallery on Your Phone
Option A: Install from Official App Stores (Recommended for Most Users)
STEP 1 Open your device's app store. On Android, open the Google Play Store. On iPhone, open the App Store. Both versions are now officially available, so you don't need to mess around with sideloading APK files unless you specifically want the latest experimental version.
STEP 2 Search for "Google AI Edge Gallery". Type the exact name into the search bar and look for the app developed by Google LLC. The app icon features a colorful geometric design. Make sure you're downloading the official app, not an impersonator or similarly named alternative.
STEP 3 Tap "Install" and wait for the download. The app itself is relatively small (around 115 MB), so it should download quickly on a standard connection. Once installed, open it immediately. There's no login screen and no "agree to terms" for data sharing, which is both odd and refreshingly reassuring.
STEP 4 Grant any requested permissions. The app may ask for storage access to save downloaded models and camera access if you plan to use the Ask Image feature. Grant these permissions as prompted to ensure full functionality.
{readalso}
Option B: Install from GitHub (For Early Adopters and Android Users)
STEP 1 Navigate to the official GitHub repository. Open your phone's browser and go to github.com/google-ai-edge/gallery. This is the official open‑source repository maintained by Google. The app is released under an Apache 2.0 license, meaning it's free for both personal and commercial use.
STEP 2 Download the latest APK file. Look for the "Releases" section on the right side of the GitHub page and tap the latest version number. Then scroll down to "Assets" and download the .apk file. The file size is approximately 115 MB.
STEP 3 Enable installation from unknown sources. Before you can install the APK, your Android device will block it by default. When you tap the downloaded file, your phone will prompt you to allow installations from "Unknown Sources" for your file manager or browser. Enable this permission, it's a standard security step that you can disable again afterward.
STEP 4 Open your file manager and install the APK. Locate the downloaded file (usually in the "Downloads" folder), tap it, and follow the on‑screen prompts to complete the installation. The process takes less than a minute. Once installed, you can open the app and begin downloading models.
Guide To Downloading and Running AI Models Offline
Once the app is installed, the real magic begins. The app doesn't come with any AI models built in, you choose exactly what you want and download it to your device. Here's how to get started with each of the main features.
AI Chat Your Offline Conversational Assistant
What It Does: AI Chat enables multi‑turn conversations with a downloaded language model, similar to chatting with ChatGPT or Gemini, but entirely offline. You can ask questions, brainstorm ideas, get writing feedback, or simply have a conversation with an AI that lives entirely on your phone. Because the model runs locally, responses appear instantly without any network latency, and your conversation history never leaves your device.
How to Set It Up: Open the AI Edge Gallery app and tap the AI Chat option from the home screen. You'll be presented with a list of compatible chat models, each with its size displayed prominently. The recommended model for most users is Gemma 4 E2B, which weighs approximately 2.58 GB and offers a text input context length of up to 32,000 tokens.
Tap the download icon next to your chosen model. The app will download the binary file directly to your phone. This may take several minutes depending on your connection speed, but you only need to do it once. After the download completes, you're ready to chat.
Before you start typing, consider enabling Thinking Mode by tapping the settings icon in the top‑right corner. This feature shows the AI's step‑by‑step reasoning process, giving you insight into how it arrives at its answers and helping you understand the model's logic.
Now, the real test: turn off your Wi‑Fi and mobile data, or enable Airplane Mode. Open AI Chat and type a question. The model will respond just as it would online, using only your phone's processor. The responses may be slightly slower than cloud AI, but the trade‑off , complete privacy and zero data usage is enormous.
Prompt Lab Summarization, Rewriting, and Code Generation
What It Does: Prompt Lab is designed for "single‑turn" tasks, meaning you provide a prompt and the model generates a response, but there's no ongoing conversation history. This is perfect for summarizing a long document, rewriting text in a different tone, generating code snippets, or exploring freeform prompts. The Prompt Lab includes several built‑in task templates and configurable settings that let you fine‑tune the model's behavior.
How to Set It Up: From the home screen, tap Prompt Lab. You'll see a list of available models, many of which overlap with the AI Chat models but are optimized specifically for single‑turn instruction following. Download your chosen model Gemma-3n-E2B-it-int4 (3.1 GB) and Gemma-3n-E4B-it-int4 (4.4 GB) are powerful options.
Once downloaded, you can access the Prompt Lab's template library, which includes options like "Summarize," "Rewrite (Formal)," "Rewrite (Casual)," "Generate Code," and "Freeform Prompt." Each template pre‑configures the model's parameters for optimal performance.
To test it, paste a long paragraph of text and select the "Summarize" template. The model will read the entire content locally and produce a concise summary. You can also use the freeform option to ask the model to write code, explain a concept, or generate creative content all entirely on your device.
Ask Image Visual Questions Without Sending Photos to the Cloud
What It Does: Ask Image lets you upload a photo from your gallery or take a new picture with your camera, then ask the AI questions about what it sees. The model can identify objects, describe scenes, read text from signs or documents, and even solve visual problems, all without ever uploading your image to a remote server.
How to Set It Up: From the home screen, tap Ask Image. The app will display a list of vision‑capable models. Look for models with "Gemma-3n" in the name, such as Gemma-3n-E2B-it-int4 or Gemma-3n-E4B-it-int4. These are multimodal and understand both text and visual information.
Download your chosen multimodal model. These files are larger Gemma-3n-E4B-it-int4 weighs about 4.4 GB, so be patient and ensure you have sufficient storage. Once the download finishes, you can start using Ask Image immediately.
Tap the camera icon or the "+" button to either take a new photo or select an existing image. After the image loads, type your question. For example: "What kind of flower is this?" or "Can you read the text on this sign?" The model will analyze the image and provide a text response based on what it sees.
Audio Scribe & AI Edge Eloquent: Offline Speech-to-Text
What It Does: Audio Scribe (built into AI Edge Gallery) allows you to upload or record audio clips and transcribe them into text or translate them into another language, all offline. Google has also released a separate dedicated dictation app called AI Edge Eloquent, which takes this capability further with real‑time transcription and automatic text refinement.
How to Set Up Audio Scribe (within AI Edge Gallery):
Open AI Edge Gallery and locate Audio Scribe. Download the associated Gemma‑based ASR model when prompted. The model size is typically smaller, making it a quick addition. Once downloaded, you can upload an existing audio file or record a new clip directly within the app. The app will process the audio locally and display the transcription.
How to Set Up AI Edge Eloquent (Standalone Dictation):
Download AI Edge Eloquent from the App Store (Android version in development). On first launch, you'll be prompted to download the Gemma ASR model. Once installed, you can start dictating immediately. As you speak, the app shows a live transcription in real time. When you pause, Eloquent automatically removes filler words like "um" and "ah," corrects interruptions, and restructures sentences to produce polished, professional text.
After transcription, you can refine the output using built‑in options like "Key Points," "Formal," "Short," or "Long." The app also tracks words‑per‑minute speed and maintains a searchable history of all dictation sessions.
Tiny Garden, A Playful Offline AI Game
What It Does: Tiny Garden is an experimental, fully offline mini‑game that uses natural language commands to let you plant, water, and harvest virtual flowers. It's a small but charming demonstration of how AI models can power interactive experiences without any internet connection. The game is powered by a fine‑tuned Function Gemma 270M model, tiny compared to full language models.
How to Set It Up: From the home screen, tap Tiny Garden. The app will prompt you to download the specialized model if you haven't already. The file is under 300 MB, so the download is quick. Once ready, you can start playing immediately. The game responds to natural language commands like "plant a rose" or "water the flowers."
It's a simple but effective demonstration of how on‑device AI can create engaging, interactive experiences that require zero cloud connectivity.
Watch Full Step by Step Setup And See It In Action
If you want to see exactly how it works before trying it, watch this quick setup tutorial. It covers installation, model download, and offline testing, step by step.
What Most People Get Wrong About Offline AI
A lot of people expect offline AI to feel exactly like cloud AI, and that's a mistake. The responses from local models may be slightly slower, and the models themselves are less knowledgeable about very recent events (since they're frozen at their training cutoff date). But the trade‑offs are enormous and, for many use cases, far outweigh the limitations.
No data usage
Once the model is downloaded, you can use it as much as you want without consuming a single megabyte of mobile data. A game‑changer for travelers or anyone on a limited plan.
No subscriptions
The app and all its models are completely free. No premium tier, no paywall, no hidden fees.
No tracking
Your prompts, images, and conversations never leave your device. Zero server logs, zero data mining.
Instant availability
On a plane, in a remote cabin, or anywhere without internet, your AI assistant is still there, ready to help.
The reality is that cloud AI will always be more powerful and more up‑to‑date, but local AI hands control back to you. As compression techniques improve and phone hardware gets more capable, the gap will continue to narrow. Within the next few years, flagship phones may ship with built‑in LLMs that rival today's mid‑tier cloud services.
The Bottom Line
Google AI Edge Gallery is a quiet revolution hiding in plain sight. It's not just a tech demo or a developer curiosity, it's a fully functional offline AI platform that anyone can use today, for free, with total privacy. Whether you're a traveler who needs an assistant on long flights, a professional handling sensitive information, or simply someone curious about the future of AI, this app is worth downloading.
The setup takes less than ten minutes. The models download once and work forever. And the peace of mind that comes from knowing your data stays on your device? That's priceless.
So go ahead: download the app, grab a model, turn off your internet, and watch your phone come alive with AI that doesn't need permission from the cloud. The future of AI isn't just in some distant data center, it's in your pocket, ready to work anytime, anywhere, without a single bar of signal.
FYI: Tutorial based on the official Google AI Edge guide and hands‑on testing.
