PocketPal AI: Android and iPhone model setup tutorial

This Italian tutorial demonstrates local model setup in PocketPal AI on Android and links to the iPhone app. The presenter installs models through the app's Hugging Face search, checks download options, and selects a model for chat. He describes offline use as a way to keep prompts on the phone rather than send them to external servers.

The walkthrough focuses on models found by searching for "uncensored". In the Models menu, the presenter taps the plus button and opens the repository search. Before downloading, he checks the app's memory warning and compares file variants. The examples include 4-bit, 6-bit and 8-bit quantization; he explains that lower bit counts reduce model size. He downloads a file described as roughly 2.5 GB. Some models require a token, and available phone memory can limit which ones he can install.

After the download, he opens Chat, chooses a model and waits for it to load. He then interrupts a response and switches to another installed model. One example produces English reasoning followed by an Italian answer. The presenter says the test prompts would receive refusals from commercial models, but the tutorial does not include a direct comparison. He stops both responses before examining them fully, so these examples do not establish answer accuracy or show that every model labelled uncensored will answer every prompt.