Layla

Layla is a private AI assistant for Android and iOS that runs local models on the device. It combines chat, character roleplay, image generation, VRM avatars, agents and 100+ voices, with an optional cloud mode.

Layla is a mobile AI assistant built around the idea that the model should live on the phone rather than on someone else's server. It runs on Android and iOS, and most of its modes process conversations locally, so chats, characters and generated images stay on the device. The app bundles several capabilities in one place: general chat, a personal assistant with on-device agents, character roleplay, image generation, VRM avatars, Python support and a library of more than 100 voices. The app opens with a choice of seven modes. Personal assistant handles everyday planning and general questions. Basic is a lightweight offline chatbot with nothing extra switched on. Friend leans toward casual conversation with memory, voice and dreams. Roleplay lets users build characters and scenarios. Image generation turns prompts into local pictures that can feed character creation. Cloud swaps in hosted models for more raw capability, at the cost of the usual provider privacy trade-offs. Expert is aimed at people who want to load their own GGUF models and control which tools and settings are active. Under the hood, Layla runs local models in GGUF (llama.cpp), LiteRT-LM and PTE (ExecuTorch) formats through a single interface. Enthusiasts can import GGUF files from any source and adjust temperature, samplers and similar parameters. A community character hub offers thousands of user-made companions that can be downloaded and kept entirely on the device, and users can also create and share their own anonymously. Much of that catalog is fiction-driven and some of it is mature in theme, which suits the roleplay audience but is worth knowing before browsing. In the wider field, Layla sits between hosted companion chat apps and hobbyist local-LLM runners. It offers more characters, voices and avatar features than a bare model launcher, and a stronger privacy position than cloud chatbots. Its natural audience is privacy-minded users, roleplay fans and tinkerers who want to run open models on a phone without a desktop setup.