RECOMMENDED PROFILE
Layla
Layla is an AI companion that runs its language models locally on the phone for offline chat, roleplay, character creation, and image generation. Most modes keep chats and inference on the device with encrypted storage, so conversations work without a network connection; an optional Cloud mode routes requests to a user-chosen hosted provider instead. Users can load their own GGUF models, attach VRM avatars, and choose from more than 100 local voices.Read full overviewCollapse overview
Layla is an AI companion that runs its language models locally on the phone for offline chat, roleplay, character creation, and image generation. Most modes keep chats and inference on the device with encrypted storage, so conversations work without a network connection; an optional Cloud mode routes requests to a user-chosen hosted provider instead. Users can load their own GGUF models, attach VRM avatars, and choose from more than 100 local voices.
Last checked2026-09-22
01 — OUR TAKE
Who and what it suits
Layla positions itself, per its official website, as an on-device AI companion: most modes run local models with encrypted chat storage, while optional Cloud mode sends requests to a chosen hosted provider. Its differentiator versus cloud-first apps is offline inference with bring-your-own GGUF models. Per its App Store listing, the iOS app is 'Layla (Cloud)', a free cloud version for less technical users, separate from the premium on-device app; exact pricing was not confirmed. Some App Store reviews report character discovery and UI confusion in the cloud app.
Best for
- People who want an AI companion that works offline without a network connection
- Enthusiasts who want to load their own GGUF models and tune generation settings
- Roleplay and character-creation fans who prefer scenarios that run on the device
- Privacy-conscious users looking for local, encrypted chat storage instead of cloud processing
Not ideal for
- Users on older or low-end phones, where local model speed and size are hardware-limited
- People who want a simple, no-configuration companion instead of enthusiast-level model settings
- Anyone unwilling to accept third-party provider terms when using the optional Cloud mode
What stands out
- On-device inference with encrypted local chat storage
- Local text-to-speech with more than 100 voices that works offline
- Bring-your-own GGUF models with temperature, sampler, and generation controls
- Community Character Hub for creating and sharing characters that run on-device
What to consider
- Layla can keep local-model chats and characters on the device, but optional Layla Cloud and third-party engines send requests off-device under different privacy terms.
- Memory quality and speed depend on the selected model, context settings and phone hardware, so results are less standardized than a single hosted service.
02 — RELATED DISCOVERY GUIDES