Player not loading? Watch on YouTube
This tutorial shows how to install Google AI Edge Gallery from the Play Store and use it to run models locally on a phone. The presenter describes the app as a mobile counterpart to Ollama, with Gemma and other open source models available to download.
The walkthrough focuses on AI chat. The model picker displays download sizes, and the presenter recommends choosing a model according to the phone's RAM and processor. He selects a small Gemma model described as roughly 500 MB, accepts the download prompts, and initializes it before opening a conversation. The tutorial also shows where to switch between downloaded models.
Chat settings include maximum tokens, temperature, the accelerator, and a system prompt. When a simple greeting produces an error, the presenter changes the accelerator from GPU to CPU and gets a response. This is a troubleshooting example from his phone, rather than a guarantee that the same change will resolve every device's errors.
The app's menu also includes image questions, audio transcription and translation, agent skills, Tiny Garden, and mobile actions. These receive a brief introduction rather than a full demonstration. The presenter notes that input support depends on the model: some accept only text, while others accept images and return text.
