New features: - visualisation of temporal relationships in Layla's long-term memory A while ago, Layla introduced temporal memories: memories that evolve in time. This update brings a whole new way to visualise your character's memories! Your memories are now laid out in a 3D space; past memories are "deeper" into the z-axis. A slider at the bottom allows you to navigate through your memories in time, jumping from past to present. You will notice that memories naturally form "clusters" in time. These are memories that occur within a relatively short timeframe. Each cluster has it's own central node(s), with auxiliary memories branching out around them. These clusters are segmented via a time-dilated algorithm: recent memories are finely-segregated, while older memories are more clustered together. This mimics how human memories work; you often remember recent events clearly, while the longer the events are in the past, the more they combine into one big cluster. Important nodes in each cluster are linked together in time. These are usually "common entities" between temporal clusters. Intuitively, this should be important characters, topics, etc. The graph reflects this. The "anchors" in each cluster are usually the two main participants of the conversation (you and your character). These nodes glow brighter in the memory network, and you can see faint "temporal links" between them. These temporal links are annotated by the time elapsed between each cluster, which is visible in the memory graph. There are a lot more ways in which this marshalled data and relationships can be used during the recall process, and this is something I will continually improve on over time! Improvements: - significant improvements to the performance of chatting, especially for very long conversations spanning hundreds/thousands of messages - long-term memory will now recall earlier memories from the same conversation that has been cycled out through context shifting - fixed a few conditions in triggering new search agent - supports the new Ministral-3 family [THINK] tags Bug fixes: - fixed a bug where invalid filename in References mini-app causes Layla to crash - fixed a bug where characters list will sometimes show nothing when updating a filter - fixed a bug where character list resets to the first page after edit/saving a character - fixed a bug where regenerate messages leaves a large amount of empty space when the length of the regenerated message varies largely - fixed a bug where automatically searching for models in OpenAI compatible API endpoints was not working - fixed a bug where Layla does not stop the server from generating messages even after pressing "stop" - fixed a bug where voice chat does not speak aloud responses - fixed a bug where audio is only played through speakers and not AirPods

New features: - Layla supports Qwen3-VL family models: images are automatically down-sized before sending to the LLM to keep processing times reasonable on mobile, you can control how much down-sizing occurs in Advanced Settings - Layla supports animated PNGs and GIFs for your character cards Improvements: - added setting in Reference documents to control the threshold of how long a document needs to be before it is summarised instead of injected wholly into the context - prompt templates are now case insensitive - display character full name in chat instead of just the first name - automatically convert JPGs to PNGs when exporting your characters as a TavernPNG - you can share text with Layla to start a chat with that text Bug fixes: - fixed a bug where the document name was not detected correctly when adding a document from chat - fixed a bug where character sorting changes back to the default after exiting chat - fixed a bug where tool calls are included in long-term memory messages - fixed a bug where Layla crashes on importing a huge TavernPNG card

- fixed bug where option toggles are not showing

- improved stability in chatting - fixed a few bugs in edit/regen - fixed bugs related to the calendar agent

Layla 6 supports Agents! Agents are fully configurable, self-contained workflows that can be triggered and executed by Layla during chats. Their functionalities can range from injecting simple context, to fully automated workflows such as reading a webpage and saving it as a reference document. Each agent is also modular: you can attach them to your own characters; they will still complete their respective tasks, but under the personality of your attached character. An agent is comprised of two parts: triggers and tools. An agent can have multiple triggers and call multiple tools. You are free to mix and match them in any combination when creating your own agents! New features: - Long-term memory has been significantly improved! - LTM now implements a 4 stage retrieval process, to balance speed and quality: - “raw memories” are created immediately as you chat; they are available immediately when starting the next conversation, no processing time needed - “embeddings” are created in the background; this process is relatively fast, and will be available 5-10 minutes after a conversation ends - “summaries” are created as a background process, when your phone is idle; this process can either be done by the built in summariser, or a chosen LLM - “knowledge graphs” are created last, with all the information available in the previous step, and require an LLM to process - During recall, memories are retrieved starting from “raw memories”, and re-ranked by content in each level as they become available, so you will immediately get memories, but the injections will increase in quality as new memories are ingested Improvements: - support for Gemma 3n models - improved prompt format slightly for vision models - backup and restore data now prevents back button presses or navigation to avoid accidentally corrupting your data if you move away from the screen - added ability configure a character level image generation prompt prefix - LTM no longer processes think tags from reasoning models - performance improvements to rendering chat messages - Layla will now automatically select the best prompt format after you've selected your model - backups & restore will now also backup your OpenAI inference settings Bug fixes: - fixed a bug where favourite characters were not being saved - fixed a bug where searching for language names in the Offline Translator app was searching for the language's English name instead of the name in the UI language - fixed a bug in native speech-to-text where it stops listening after the first round - fixed a bug where horoscopes were not showing - fixed a bug where a character with an empty description stops all characters from showing up - fixed a bug where OpenAI inference settings was not backed up

- fixed bug where Metal GPU inference was not working

Improvements: - added support for Kroko ASR (speech-to-text model, this is now the default STT model in Layla) - ability to configure KV Cache quantisation settings in Advanced Settings - added French, Portuguese, and Spanish UI translations Bug fixes: - fixed bug where the correct number of characters in Roleplay was not showing - fixed bug where chat images cannot be expanded when tapped on - fixed bug where slightly moving your finger during hold-to-speak mode stops Layla from listening - fixed bug where sometimes phone native voices are not being shown or not working in text-to-speech

- fixed bug where edit and delete actions were not showing for previous messages - removed animated backgrounds temporarily due to a bug

New features - app UI has been translated to multiple languages! Currently supports: English, Chinese, Korean, Polish, Russian - Offline Transcription: transcribe audio such as lectures, audiobooks, or videos in real time - all preset characters now have animated backgrounds! Improvements: - allow configuring minimum silence length and other settings for speech-to-text - made stop generating slightly more responsive - improved UI performance when generating responses, should reduce speed inconsistencies where response times sometimes start slow and speeds up at the end of the message - added TopK in advanced settings Bug fixes: - fixed bug where DuckDuckGo news search was not working - fixed bug where multiple regens make the model increasing incoherent - fixed bug where text-to-speech sometimes pauses unnecessarily after the first sentence

- fixed bug where the AI sometimes repeats what you said before responding

- fixed a small bug where chat screen is not redirecting to Layla Cloud login if you are logged out

New features: - Gemma3 family models support! - added new CPU stable diffusion models Improvements: - Layla supports displaying LaTeX formulae in chat messages - backup data will now backup your custom prompts - added ability to copy generated code in chat messages - added support for MultiTTS voices - added ability to continue generating a message from the LLM Bug fixes: - fixed bug where GPU inference was not working with Background Inference

New features: - added a new animated character - added ability to duplicate characters (including preset characters and animated characters, which will allow you to use the animation for your own characters) - added XTC samplers in Advanced Settings - added support for Phi-4 family models Improvements: - added feature to generate image prompt using your selected LLM in SD mini-app - improved animated character lips movement; lips will start moving when sound is played instead of on LLM output Bug fixes: - fixed bug where saving backup data directly to file system does not work for very large backups - fixed bug where character info is not updated after editing until a reload - fixed bug where deleting a character does not refresh the characters list - fixed bug where TTS voices were not playing in phone silent mode - fixed bug where TTS voices do not play after stopping it once - fixed bug in image generation resulting in "workspace alignment" issues - fixed bug where character tags are not showing in the character creator - fixed bug where deleted characters were still sending proactive messages

New features: - added iniVerse Mix Layla Cloud model - added Cohere API support - added support for importing webpages as Lorebooks - added support for Kokoro TTS model Improvements: - optimised Character Selection page: added sorting options, optimised loading a large number of characters Bug fixes: - fixed bug where {{instruction}} tags were being force injected for some characters

New features: - Layla supports reasoning models Deepseek R1 family! Improvements: - redesigned Lorebook UI to handle lots of documents better - improved UI of model import - added timestamps to Long-term Memory table view - backup data now directly allows you to choose a folder to save to - added a Download Manager app to give the ability to view/cancel download tasks in case they get stuck - added Whisper Base and Whisper Base (English) models - added ability to configure the language Whisper models listen in - Q4_0 quants are now automatically converted on the fly to support your current architecture - allows saving TavernPNG directly to file system in character creation - supports sherpa-onnx TTS engine APK - redesigned chat message quick actions (copy button is now always visible, tap & hold the message to bring up a context menu with more action) - Create Character (AI) image generation now uses the default negative prompt configured in the SD mini-app Bug fixes: - fixed bug when importing chat history - fixed bug in Layla Cloud when handling very long conversation histories - fixed bug where an error in one memory will stop ingestion of all LTM memories - fixed bug where too many quick actions take up all your screen in chat - fixed bug where chat accent colour was not being applied to character responses - fixed bug in default character image generation fallback phrase

- fixed bug where sometimes LTM says task not found - fixed bug where sometimes messages were not being sent properly to Layla Cloud

Improvements: - default characters can now generate images based on your Inference Settings Bug fixes: - fixed a bug where Diagnostics was not working - fixed bug "zip headers not found" when backing up data - fixed bug where Lorebooks overall summary injection was not working - fixed bug where Lorebooks max injection number setting was not working - fixed a bug where multiple same memories are being injected into the same conversation - fixed a bug where multiple memories were not being injected

New features: **Layla supports local image generation via Stable Diffusion!** - configurable prompt/negative prompt, step number, and guidance scale - ability to connect to your local instance of Automatic1111/stable-diffusion-webui on your PC - Layla Cloud supports image generation _Characters can now generate images during chat!_ - added new settings in character creation which allows you to set per character image generation settings - configurable fallback phrases to force your character to send you an image - quick message action to generate an image using the contents of a particular chat message **Model Benchmarks app - benchmark your local LLMs on your phone** - run inference on local LLMs, store tps, ram usage, cpu usage, etc. in a table - generate report with hardware info and benchmark results **AI Character Creation** - generate characters by inputting keywords or vague descriptions - LLM will generate the character description and personality for you - generate images for your character using Layla Stable Diffusion **Long term Memory** - re-wrote Long-term memory with significant algorithmic improvements - employs a 3 stage retrieval algorithm to increase recall precision dramatically - LTM now supports "temporal recall" (you can mention something like *"that conversation we had 3 months ago"*, your characters will be able to pinpoint the conversation summary at that time, if any) - memories are now strengthened over time, earlier memories are now less likely to be returned, however, frequently recalled memories will offset this **Welcome Screen redesign** - Layla v5 comes with a completely new on-boarding experience - Choose from 6 different uses for your on-device AI: - Basic - LLM only, fast, clean, no-frills - Assistant - professional, tool-calling, good at general knowledge - Friend - empathetic, casual, with long-term memory - Roleplay - multi-characters, Lorebooks, character creation - Image Generation - on device image generation with stable diffusion, AI powered character creation - Cloud - access to a wide variety of powerful cloud models, limited privacy - Expert - bring your own models, enable/disable features of your own choosing Improvements: - all mini-apps can now be downloaded in the background - added configurable Lorebook injection prefix and suffix - redesigned Personality Hub, Character Selection screens - added ability to load GGUF models from external SD card - added advanced setting "max response length" - added import/export chat history feature - added UI settings for character text colour - added ability to select multiple messages in chat - selecting the "Default" preset now resets all important settings in the Advanced Settings to their defaults (not just sampler settings) - added setting to toggle whether to "tap and hold message to copy" - redesigned character tag filtering in Character Selection and Personality Hub - added ability to share debug logs of individual chats - added a special "Cloud" prompt format Bug fixes: - fixed bug where greeting was not being imported from TavernPNG - fixed bug where sometimes cloud models do not respond with anything - fixed bug where regen was removing an extra token, causing the regenerated response to be unpredictable - fixed a bug where a chat was "locked" into tool calling mode after the first tool call

- fixed bug where roleplay app was not working - fixed bug where to-do app was not working - synced llama.cpp to latest version

New Features: - Tasker Integration! Layla will show up as a Tasker Plugin Improvements: - allow selection of character greetings before starting the chat - added "basic characters" roleplay mode, where you can add short descriptions without creating a separate character - added support for SmartVoice TTS: https://poretsky.github.io/android/smartvoice/index.html - Roleplay now supports Background Inference - Voice Chat app now allows you to configure different Whisper models for speech-to-text - improved "stop generating" button to be able to stop even while processing input Bug fixes: - fixed bug where context length setting was getting reset every time you switch models - fixed bug where chat name was not being saved - fixed bug where things get messed up if you exit a chat before waiting for it to load

New features: - Layla now supports SillyTavern Lorebooks! You can import them as a document in the Layla Lorebook app - SillyTavern Lorebooks work differently than normal documents - instead of using similarity embeddings, SillyTavern Lorebooks are activated via keywords (for more info please read the SillyTavern lorebook documentation on their website) - Layla's implementation retains this functionality, but adds similarity embeddings _in addition_ to the regular SillyTavern activation logic - you can create your own lorebook in SillyTavern on the PC, or download lores other users shared here: https://chub.ai/search - added "Push to talk" mode Improvements: - llama.cpp updated to support Nvidia's distilled Llama3.1 models (https://huggingface.co/nvidia/Llama-3.1-Minitron-4B-Width-Base) - added setting to turn off the dark image overlay when showing NSFW images in the Personalities Hub - Roleplay app now supports attaching Lorebooks! - allow using camera to send images in chats - allow adding Lorebooks to chat on the fly - added ability to import character expressions as zip files - added "keep scroll at bottom during chats" setting Bug fixes: - fixed bug where ePub files were not importing correctly with spaces in the document name - fixed bug where long-term memories were not being deleted for a particular character - fixed bug where a long first message does not scroll - fixed bug where Knowledge Graph construction was not respecting the LTM settings - fixed bug when selecting character expressions

New feature: - Roleplay app has been updated to v2: introducing Group Chats! - create scenarios and add your characters as participants - choose between different levels of interactivity: - None: don't interact with the group chat, your characters will continue the conversation/roleplay on their own, while you sit back and watch them - User: join the roleplay as a regular participant, you will be prompted for your input at regular intervals - Narrator: join the roleplay as an omniscient narrator, your input will be wrapped with special [scene] tags in order to advance the scenario, you can choose between allowing characters to continue the scene or narrating changes after every message Improvements: - character selection screen grid view now shows recently chatted characters - Executorch now supports Long-term memory knowledge graph construction and character impressions - Executorch now supports Dream - Executorch now Layla's Toolkit - synced llama.cpp with latest, merges in changes to support Llama3.1 and Gemma models - Lorebooks are not properly backed-up - added viewable chat information for each chat, including cache file size and which model was used to create it - improved Dream scheduling of messages to avoid spam - added Apple OpenELM and Gemma2 2B prompt settings - Offline Translator text-to-speech now speaks the translated text instead of the original English LLM output Bug fixes: - fixed bug during TavernPNG import with special characters - fixed bug where editing personas were overwriting the default persona - fixed bug where chat history was not opening in character selection screen - fixed bug where going back from the chat screen is not opening the messages screen if you have your start screen set to "messages" - fixed bug where the UI gets confused when swiping in memory cluster view and opens the chat history - fixed bug where quick actions were not showing - fixed bug with Lorebook app processing very large PDFs - fixed bug where small app icon layout was bugged

New features: - Layla now supports Personas! - personas are descriptions of "yourself" that characters know before starting a conversation - create/edit different personas when chatting with your characters, easily switch between them in Inference Settings - integrated with Custom Engines, attach personas to characters to automatically activate them when starting a chat - attached personas will also affect their Long-Term Memories and Dreams - download individual characters as TavernPNG to share privately with your friends Improvements: - improved chat history display, it is now always accessible by swiping from the left similar to other apps like ChatGPT Bug fixes: - fixed bug where some icons are showing up as question marks - fixed bug where Dream messages were disappearing after reading

New features: - Offline translator now supports "English - Spanish" and "English - Chinese" - ElevenLabs text-to-speech support! - Layla's Toolkit has been upgraded to v2! much more accurate tool determination, much faster tool determination (5s - 0.5s) - Device calendar integration for scheduling events (uses your default calendar set in your system settings, can be Apple Calendar etc.) - Todo app is integrated into Layla's toolkit as a calendar provider Improvements: - added setting to stop running background tasks when phone is idle - create multiple alternative greetings for a character (greetings are now optional, you can choose to not have a greeting for a character) - added character names in the LTM app when viewing memories - added ability to reset character impressions Bug fixes: - fixed invalid Executorch models download link - fixed bug where preset characters were not hidden in the chat history screen - fixed bug where roleplay images were not saving - fixed bug where some messages were deleted after editing or regenerating messages - fixed bug where Dream messages revert back to the character's original greeting

New features: - PDF Digest has been redesigned -> it is now called Lorebook! - process documents and attach them to characters to provide background information during your chats - Added support for the open source search engine: Alexandria Improvements: - improved performance of task manager when showing a large amount of pending tasks - added configuration to off-load embedding service to the CPU to save RAM - Voice chat supports third-party TTS libraries RHVoice and AcapelaTTS - added support for native phone speech-to-text - added ability to copy translated text in Offline Translator - added config to set minimum number of characters to generate before starting TTS - added button to "hold to continue speaking" in voice chat mode - improved UI of Offline Translator (show downloaded models at the top) - allow deletion of downloaded translator models - Offline Translator now remembers your last selected language - added ability to configure whether to use mmap in advanced settings Bug fixes: - fixed bug where tasks are not deleted after deleting a Lorebook PDF - fixed bug where sometimes new messages are duplicated as they are received - fixed bug where re-entering voice chat mode from within the same chat does not start listening