In this update, we introduce an important feature: local image recognition!
Previously, image recognition was done on the server, it has now been moved to your phone! Using the latest MobileVLM model, which stands for Mobile Vision Language Model, you are able to chat with your images.
Characters in Layla also understands images and can chat with you naturally when you send an image. For example, sending a picture of good food will have Annie comment on how good it is!
Full Changelog
New features:
New app: MobileVLM - Layla can now recognise images on device, no more sending your images to the server!
Improvements:
if Layla exits unexpectedly during chat, your last message is automatically set into the text box when you load the conversation history
backend stability and performance improvements
added ability to duplicate chat sessions
Bug fixes:
fixed bug where if you quit chat unexpectedly, the last message you sent is displayed, causing you to be unable to edit or regenerate last response
fixed bug where context shifting was not adjusting the session properly as conversation approaches the model's context length
Comments