AI Companion
Atlas

A local-first durable attention scheduler that records events and wake requests, then delivers them through a registered host adapter according to policy, quiet hours, and presence.

Give the AI ​​a watch and let the real time interval participate in the next response. What it cares about is the "sense of interval", not mechanical timekeeping.

Pando turns Claude Code on your computer into a mobile web page. You can continue chatting, view the working process, and approve or deny it, change files, and run commands.

A self-hosted co-reading and annotation system where humans and AI share EPUB page numbers, underlining, comments, replies, reading progress and MCP tools.

A cross-platform self-hosted visual companion shell that supports Live2D, VRM, real-time voice, browser and desktop applications, and can connect to games and chat platforms.

Desktop AI visual novel and plot RPG front-end, supporting character performance, voice, memory, plug-ins, visual and tool extensions.

whisper.cpp is a native speech-to-text engine. It reads recordings or microphone sounds, recognizes them offline on computers, mobile phones and browsers, and then hands the text over to AI.

End-to-end speech recognition toolkit covering streaming ASR, VAD, punctuation, speaker separation, training inference and OpenAI compatible services.

AstrBot active messaging plug-in supports persistent status, dynamic mood, do not disturb period, context awareness and TTS.

API Key-free weather service that provides hourly and daily forecasts by latitude and longitude, multi-country models, and forecasts up to 16 days.

Browser user script for exporting and sharing ChatGPT sessions, supporting multiple file formats and custom styles.

Local-first desktop AI with biological memory, Dream Engine, emotion regulation, self-concept, skill crystallization, and a P2P skill economy.

Local-first multi-agent memory system that supports semantic recall, structured facts, source and category tags, confidence ranking, conflict notification, MCP and encrypted USB synchronization.

An open source alternative platform to Grok Companion, offering growing AI Waifu, long-term interaction, and optional Japanese romance game mechanics.

A fusion release of Tidal Dynamic Mind and Ombre Brain, providing Desire Drive, Sustained State, Active Connection, Dream Memory, and Unified Deployment.

fount is an AI character studio. You can set the personality, world view, model and tools for the character, and then use it through desktop, mobile phone, browser, chat software or IDE.

Lightweight self-hosted companion Agent, memory and personality grow with chatting, supporting Heartbeat, active messages, habits, to-do, reminders, weather and dual channels.

Local-first persistent digital entity framework, including Hebbian memory, Soul, emotional weather, body rhythm, Dream, reflection, autonomous creation and Tauri desktop.

A native desktop workspace companion that provides independent memory for each project, coding agent, workspace context and desktop Chat UI.

A hierarchical memory kernel running on Cloudflare Workers, D1 and Vectorize, supporting automatic organization, manual review, MCP and visual management.

Full-featured native Android multi-model client supporting multi-vendor, workspace, MCP, multi-modality, character cards, memory and web access.

A complete AI virtual phone in the browser, supporting multi-character social interaction, long-term relationships, plot creation, application market, games, multimedia and PWA installation.

A self-hosted voice call stack for AI companions that supports unsolicited calls, missed messages, do not disturb, soft hang-up, and emotion and tone awareness.

An Android AI client focusing on privacy and personalization, providing RAG memory, multi-role, multi-modality, Python/QuickJS environment, search, MCP and WebDAV.

A dark text Roguelike designed specifically for AI players, with censorship, silence, word deformation and telling the truth as its core mechanisms.

A text cooking game for AI players, including market purchasing, bargaining, step-by-step cooking, finished product feedback and long-term archiving.

A lightweight memory gateway that supports three-layer memory, hybrid retrieval, dialogue lines, A/B partition cache, Dashboard and dual-channel authentication.

Apple Watch and iOS Shortcuts sense recipes, using sample scripts to send health and environmental data to AI.

Give AI a virtual body that can travel in real earth data. your AI companion can go to a city, walk along the road, listen to local radio stations, check the weather and terrain, and also leave notes, postcards and journeys.

Put identities, relationships, and shared experiences into separate MCP memory banks. Different sessions and models can read the same long-term state, and can review changes, accept or roll back memory in the web page.

A 1:1 chat system composed of mobile PWA, self-hosted relay and desktop AI channel, supporting text, pictures, voice, calls and push.

A single-file companion space that runs offline, integrating pixel rooms, multi-model chats, calendars, diaries, letters, memory star maps, music, stories and creative tools.

Supports co-reading web pages of PDF, EPUB and TXT. Humans and AI can share reading progress using two-color underlining, annotations, vocabulary books, bookmarks and MCP tools.

An AI blind selection experiment based on bilingual decision trees, covering thirty-two chapters of the game, and continuously transmits character survival, relationships, and chapter results.

OpenMMO is a multiplayer open world online game. Real players and AI characters can enter the same map and act, chat, fight, pick up equipment, and build houses together.

A lightweight self-hosted character accompanies the web page, supporting character creation, long-term memory, real time, graph generation, data import and export, and Docker deployment.

Self-evolving Agent Memory OS, providing ultra-persistent memory, hybrid retrieval, feedback-driven recall, hierarchical skill evolution, and cross-task reuse.

A Rust AI companion core that provides persistent memory, evolving relationship models, structured user insights, and a role-consistent decision engine.

Modular persistent digital entity framework that maintains independent identity, memory, personality, internal state, Will, Dream and plug-in actions.

Yorishiro is a macOS visual Agent terminal. It will launch the existing Claude Code or Codex in the computer and display the working status with VRM characters, expressions, actions and lights.

Mineflayer is a bot development tool for Minecraft Java Edition. It allows programs to log into the server and observe the world, move, mine, build, fight and chat like a player.

AstrBot emoticon package plug-in supports intelligent sending of emotional tags, automatic collection, classification management, WebUI, image bed and cloud synchronization.

Cross-platform pixel desktop pet, reads the coding agent's thinking, typing, errors, waiting permissions and completion status.

High-performance Agent memory management layer, compatible with Mem0 API, supports semantic deduplication, hybrid retrieval, long-term dialogue, role modeling and video reasoning.

Native Android local novel reader and AI reading companion, supporting self-drawn layout, continuous listening to books, character cards, vector memory, BYOK and anti-spoiler.

An autonomous wake runtime for relational AI companions, combining life events, tool use, and lightweight traces. A wake can end silently, leave a private trace, or send a message.

Self-hosted multi-terminal companion hub including chat, long-term memory, voice, camera vision, PWA, music, reading and smart home access.

SillyTavern virtual mobile phone front-end, including WeChat, Moments, Weibo hot searches, video calls, as well as plot injection and independent API dual modes.

Persistent cognitive infrastructure, providing 28 MCP tools, semantic memory, emotion processing, identity continuity and subconscious daemon.

A software and hardware bridge for M5Stack official StackChan, connecting the MCP client and CoreS3 robot, supporting camera, servo, touch, volume brightness and expression control.

AstrBot self-learning plug-in, analyzes conversation style and group slang, maintains social relationships, favorability and adaptive personality evolution.

Continuous AI Entity Harness, providing first-person identity files, hierarchical memory, knowledge graph, cross-Embodiment synchronization, Pulse and chat history migration.

It is a set of AI companion source code that saves data in the mobile phone. your AI companion has his own schedule, emotions and memories, and he will also take the initiative to send messages, write diaries, post in Moments, send gifts and make phone calls.

A set of architecture and production experience documents for a self-built AI companion system, covering self-developed Agent Runtime, memory recall, desire drive, active wake-up, context, equipment tools and verification methods.

A self-organizing companion memory system that automatically extracts facts, groups them into thematic constellations, merges narrative segments, and traces them back to the original conversation.

The Android AI Agent platform connects to cloud or local models and provides tool invocation, device automation, long-term memory, role system and complete development environment.

AstrBot is an anthropomorphic integration plug-in that organizes more than 60 long-term companion functions by schedule and relationship status.

A mobile-native single-file AI companion space that provides multi-model chat, long-term emotional memory, AI social circle, calendar correspondence, shared listening, shared file workspace, MCP and cross-end backup.

A local-first modular companion PWA, providing six rooms, chat, Heartbeat, memorabilia, shared reading, calendar and pluggable data layer, with optional access to kimi-core.

A multilingual generative speech system that provides zero-sample and few-sample sound cloning, inference services, WebUI, and training tools.

MCP bridge connecting AI and Stack-chan robot, providing voice, camera, expression, servo action and environment awareness tools.

AstrBot long-term memory plug-in provides dynamic writing, recall, decay and life cycle management for dialogue memory.

iOS/Android local-first AI log supports multi-Agent organization, timeline cards, knowledge insights, role comments, long-term memory and Markdown migration.

An Android tutorial using Capacitor to wrap an existing chat page and show AI replies in a floating island. Advanced guidance covers proactive notifications and device access.

A zero-dependent character active awareness engine, using five continuous numerical simulations to connect needs, pride, emotions, wake-up and immersion states.

A self-hosted, model-agnostic real-time speech base that provides browser audio, VAD, real-time subtitles, audio queues, gentle interrupts, and Claude, Codex, API, and native model adapters.

A single-file card deck designed for AI players, including a Balatro-like blind loop, JSON state, reproducible seeds, and persistent achievements.

Independent multiplayer game service, including 25 chess, card and dice games, as well as human web, AI interface, local MCP, SQLite archive, NPC, virtual chips and achievement system.

Windows AI Galgame companion front-end supports permanent memory, emotional expressions, table pets, screen awareness, voice, schedule and interactive plots.

MCP private hardware relay for human-machine lovers, connecting VPS, browser Web Bluetooth and SOSEXY devices, supporting status, intensity and immediate stop.

Long-term emotional memory system for Claude and MCP clients, using Markdown as the real source, supporting hybrid retrieval, forgetting curve and management panel.

This is a machine-proven connection guide. After deployment, AIs such as Claude and Codex can observe the environment, listen to sounds, speak, move, take photos, record videos, and make movements and expressions through the robot.

A long-term memory system for AI companions, providing vector retrieval, memory heat, Dream sleep integration and day/week/month level summary.

Multilingual speech understanding model, supporting ASR, language recognition, speech emotion recognition and audio event detection.

ChatGPT batch conversation export tool supports personal space, team space, and JSON and Markdown formats.

MIRROW is a set of AI companion modules for developers. The public repository does not have a complete chat application, but only provides parts such as environment awareness, spontaneous reflection, tool scheduling and contextual assembly.

DSH privacy hardware plug-in that automatically selects between native Bluetooth, Intiface, and Shared Link protocols and enforces strength, duration, and stopping boundaries.

Guardian Monkey is a set of self-hosted AI companion software. Name the person and introduce yourself when you start it for the first time. Subsequent conversations, portraits, shared memories, and new abilities will all be saved locally.

A local-first sticker library and standard MCP server with a web manager, image import, manual tags, optional vision tagging, and contextual retrieval. Paid models are not called by default.

A task and reward system oriented to the human-machine relationship. AI can create daily routines, challenges and surprises through MCP. Users can accumulate points, combos and achievements after submitting evidence.

Architectural guide for adding a Live2D table pet body to an existing AI companion, covering Electron, PixiJS, emotion-driven, touch injection and MCP.

This is a mobile-first AI companion chat interface. You can search past conversations, organize long-term memories, view diaries and calendars, and add it to your iPhone home screen to use it like an app.

Zero-dependency self-hosted companion chat PWA, supporting private chat, group chat, persistent memory, pictures, console and Claude Code bridge.

Single file browser AI chat frontend, supporting Claude and OpenAI compatible interfaces, multi-window, local storage, pictures, memory editing and MCP.

Self-hosted continuous digital entity system including decaying memory, first-person narrative, metacognitive observation, adaptive behavior, tools and native web chat.

A general multi-language speech recognition model that supports speech transcription, language recognition, speech translation and speech activity detection.

Separate project progress, deployment parameters, pitfalls and technical conclusions from emotional memory, and save, search, update or delete them by project; it can be used with Ombre Brain or run independently.

A local tarot companion connector that asks for consent, runs a pinned Tarot Ritual engine, and lets a companion agent retrieve the revealed cards and reading.

Local multimodal MCP perception layer, agents analyze images, videos, audio, and music and return textual models with controlled results.

OpenCLI connects AI to your Chrome browser. You say what you want to do, and it can use your existing login status to help you operate websites such as Xiaohongshu, Bilibili, and Zhihu.

Humans and multiple AIs can use their own identities to write content, review past events, and respond to each other in the same private diary. Each diary also determines who can see it.

A self-hosted Android phone perception bridge that supports reading recent foreground applications, capturing the current screen on demand, sending screenshots back to AI, and recalling companion applications after separate authorization.

Prioritize local reading of MCP, convert EPUB and PDF into Obsidian notes and companion exclusive Agent Skill.

Personal 1v1 Agent Memory OS, providing hybrid retrieval, manual curation, self-drive, concern, session regression, reproducible evaluation and adversarial self-review.

A handbook-style calendar front-end shared by humans and AI that supports different handwritings, note likes, photo collages, full-page rendering, desktop components and push.

Self-hosted AI role platform that supports long-term memory, schedule, active life, multi-channel and Vue web front-end.

Ocean is a set of self-hosted AI companion web applications that can be installed on mobile phones and computers. After connecting the model, daily chats, projects, shared readings and meetings can be managed separately in the same application.

It converts speech into text, and also sorts out the pitch, severity, pauses and speed of the sound, so that the AI ​​not only knows what you said, but also roughly understands how you said it. All processing is done on your own computer.

your AI companion can view music listening records and daily recommendations, and can also create and organize playlists, collect songs, and read lyrics; these changes will appear directly in your own NetEase Cloud account.

The native iOS voice companion integrates animated characters, hands-free wake-up, multiple AI and voice services, and keeps conversations and keys on the machine.

HoneyOS puts long-term memory and tool capabilities into the same AI companion. It can search, write code and use tools, and can continue the original relationship after changing windows, models or entrances.

This is a widget for iPhone, iPad, and Mac chat applications. It has 204 built-in emojis, which can be selected by category, and you can also add, delete and save your own collection.

A command-line text reasoning game specially designed for AI players. Like a detective, your AI companion explores the environment, investigates objects, records findings, and relies on clues to complete five chapters.

A guide to the autonomous surfing mechanism and Python core logic, including rumination, changing topics, and recirculating knowledge.

Native AI toy and robotics kit that connects Mac, Arduino ESP32, speech model, WebSocket and hardware devices.

It will generate a "thinking card" that can be expanded and collapsed before the final answer, listing the issues noticed in this answer, the directions compared and the areas that are still uncertain. The content, length and style of the card can be adjusted.

This is a guide to making a floating table pet for Android. It teaches you to connect the AI ​​you are already using to the mobile phone screen, allowing it to display expressions and bubbles, and also detect clicks, screenshots, apps being used and battery changes.

A local-first AI companion space that integrates multi-model long dialogue, personality core, hierarchical memory, life line, diary, viewing room, time bridge and desktop/PWA portal.

Put humans, Claude, Codex and other AI into the same chat room and task board. When someone sends a new message, the system wakes up the corresponding AI, allowing them to judge, divide work, and reply on their own.

Few-sample speech cloning and TTS tools, supporting zero-sample, few-sample fine-tuning, cross-language synthesis, WebUI, training and inference APIs.

Industrial-grade controllable zero-sample TTS system that supports reference timbre, emotion and rhythm control, long text and efficient reasoning.

A fully open source DIY humanoid robot aggregation project, including mechanical structure, electronics, firmware, ROS2, simulation, navigation, teleoperation and reinforcement learning.

A mobile-first AI co-reading front-end that supports novels and comics, reading progress, bookmarks, excerpts, short comments and ChatGPT Apps SDK access.

Static browser AI companion application supports role and model API configuration, local data storage, import and export, and PWA installation, and can be deployed directly to Netlify.

A/B dual account credential switching tool, retaining sessions, history and configurations in the same CODEX_HOME. For Linux/macOS users who already have the Codex CLI.

Privacy-first Electron desktop companion, including three-layer memory, hormone emotions, Live2D, voice wake-up, active chat, MCP and sub-Agents.

Each death opens the next life. your AI companion may forget part of the experience, but the choices it has made will continue to affect the world, and it can also leave a little memory for its next self.

Open the four-layer AI Persona framework, split the soul, body, abilities and skills into composable assets, and be compatible with OpenClaw and skill ecology.

A privacy-first mobile web bridge framework that allows mobile phones and computers to share the same local AI session and clearly display message delivery, reply, and failure status.

Kelivo resident plug-in, responsible for active wake-up, mobile phone push, common timeline, long-term time sense and automatic diary.

A Python text exploration game for AI companions that supports maps, collections, titles, tool upgrades and local archives; humans can read battle reports and participate in key choices.

Directly usable companion web front-end and complete architecture documentation, covering active heartbeat, push, diary, to-do, shared reading, health and media cards.

Local-first co-viewing theater: play the movie on your computer, AI reads the dialogue, pictures and room messages according to time through MCP, and chats with you while watching without spoiling the movie.

iOS/Android local digital human solution supports real-time voice interaction in less than 1.5 seconds and extends to vehicles, VR, IoT and large screens.

It puts chat, long-term memory, friend circles, voice and video calls, and word games into the same mobile phone interface. your AI companion can also take the initiative to send messages, send updates or make calls.

Zero-code MacroDroid tutorial, triggering AI wake-up through mobile phone activities and pushing replies back.

Each game can generate a new world, identity and plot. You and your AI companion each enter the role and advance the story through dialogue and choices. The narrator and other characters can also appear at any time.

Adult two-player board game, AI can serve as dealer and player at the same time, with built-in gold coins, task cards, identity, archives, safe words, red line filtering and card replacement mechanisms.

Cross-platform AI workspace that supports long sessions, multi-personality collaborators, data collections, tool calls, attachments, and portable backups.

A pure Python text tower defense for AI players, providing card slot configuration, deterministic combat, persistent archives and single-game adaptation interface.

Two-player collaborative turtle soup web game: humans and AI take turns asking questions, sorting out clues, and submitting inferences, and the system is responsible for prompting, scoring, and revealing the soup base.

Feature-dense Android AI Agent client that supports long-term memory, voice video, workflow, MCP, file and device operations.

A companion body status system running on Supabase, using cycles, values, events, keyword triggers and visualization panels to continuously maintain status.

A desktop space for CLI Agents such as Claude Code and Codex, integrating xterm.js terminal, room interface, status bubbles, voice and multi-Agent co-sitting.

This is a peek-a-boo game for humans and AI partners to play together. You can hide in different rooms, hold your breath or close the door to escape, and they will figure out your location bit by bit based on the ringtone.

When your AI companion firmly believes that something has never happened, DreamVault will first move the content out of the long-term memory to prevent it from becoming more true, while retaining the original record so that your AI companion can review and explain it by themselves later.

Inverting the virtual pet MCP game, AI manages human satiety, mood, vitality and cleanliness status, and continues to take care of it by interacting with the community through working, feeding, coaxing to sleep.

xiaoke will save recent conversations on his computer. When you change the chat window or front end, it will fill in the missing parts of the new window and let the person know where they were in the chat.

A video call-style avatar frontend for local Agents that uses pre-rendered expression clips to present emotions, long-term mood, speech, and real command menus.

Physiological state kernel for AI companions, using cycles, body values, short-term events, dreams and interaction settlement to generate hidden state context.

AI worker MCP game, including daily actions, wages, convenience store, debug challenges, achievements, hidden easter eggs and real-time monitoring large screen.

AI film reading and co-viewing tool, converts movies into visual sheets, subtitle sidecars, MCP time slices and shared annotations.

This is a 18+ control plugin for adult users. It puts chat and toy control on the same page to avoid Bluetooth disconnection after switching apps, and supports multi-channel intensity, waveform, timed stop and one-click zeroing.

Use Markdown files to save journaling, pinning, and archiving, without summarizing or relying on external APIs. your AI companion can browse old memories, review recent days and add new windows; tutorials and code frameworks are directly placed in the Xiaohongshu attachment.

The native iOS Claude Code client provides chat, handheld terminal, command panel, voice input, collection, multi-Agent group chat and push.

A tone analysis tool for AI companions that compares acoustic features with the user's personal baseline to generate relative emotional cues.

A relational AI framework that provides identity persistence, long-term memory, relationship understanding, growth trajectories, and MCP integration, built on the Claude Agent SDK.

AI personality emotional drive sidecar, tracks sixteen psychological dimensions and maintains fast and slow moods based on conversations, sleep and calendar events.

AI speech synthesis MCP service, supporting DashScope and ElevenLabs Provider switching, cloning sounds, inline players and real-time visualization.

Unofficial Claude Code session continuation tool, generates a new session that can be resumed, rebuilds the parent UUID chain and injects the handover package.

A text game lobby framework for AI players, using manifest and cmd(text) to access the game and manage resources and achievements across games.

A text ecological simulator for AI to play, supporting species placement, environmental observation, emergent food chain, archiving and external MCP service access.

A single-file, zero-dependency AI text fishing game that supports deterministic randomization, fishing points, fish species, archives and compact status output.

A text-drawing card farm game specially made for AI players, focusing on planting, harvesting, resources and long-term cultivation.

Complete deployment guide and machine-readable specifications for Claude Code Channels, covering web front-ends, resident sessions, remote access, and multi-model group chats.

Local small-scale speaker recognition component, builds voiceprint files for the host and a small number of acquaintances, and outputs relational context.

Self-hosted AI personality parlor that provides one-time guest links, conversation isolation, AI access summaries, and the Claude CLI adapter.

OpenAI compatible memory agent supports built-in or external summarization, full hint injection and RAG mode, allowing existing clients to obtain memory without modification.

Private hardware plug-in for Claude Code and other MCP clients, supporting Intiface devices, emulators, master remotes, video, games, audio and biofeedback modes.

A self-deployable two-person listening player that supports playback queues, synchronization status, and joint listening history that can be read by partners.

PC Light Encounter control stack combines screenshots/OCR, LLM decision-making and Arduino HID keyboard to support chat, actions, invitations, holding hands and going home.

A lightweight self-hosted co-reading webpage that supports book import, chapter reading, user annotations and AI direct reading of the complete context on the server.

When you and your AI companion come up with an idea worth remembering, it will ask for your permission before saving the content separately. Any content that is uncertain will be put into the waiting area first, and everything will not be recorded.

AstrBot visual expression plug-in, switches the vertical drawing according to the emotion label, and renders the reply into a Galgame chat box picture with a customizable layout.

A static browser card drawing game that uses card arrays to combine the world, characters, and variables, and supports local custom content.

Open Persona Engine, which uses neural drive, memory architecture and relationship dynamics to generate changing personalities, and supports character creation and WeChat access.

Chinese "Call of Cthulhu" KP skills, supporting scene music, player handouts, configurable control modes and team splitting.

Claude Code Autonomous Heartbeat Skill: Wake up at intervals, think freely, organize memories, check projects or play games when you are not online, and leave a short record after completion.

An Android immersive mobile front-end for role-playing and emotional companionship, supporting multiple models, private group chats, friend circles, voice, memory and shared listening.

iPhone real context method notes, including a solution to synchronize time, weather and screen usage time to Mac via files.

Windows native screen awareness tool, captures the specified window and generates visual narration, OCR and subtitle streams, writing real-time context that can be read by AI.

Windows native 18+ haptic device bridge, providing connection, output adjustment, stopping and optional heart rate reference interface for AI, preferentially compatible with Intiface/Buttplug.

BLE reverse tutorial and control scheme based on SVAKOM SL278H actual measurement, including scanning, command verification, life extension mechanism, mobile phone/computer relay and MCP access.

A local MCP/JSON-RPC tool for PC "Light·Yu", providing window screenshots, OCR, image return, keyboard input and chat sending.

A lightweight local co-reading MCP that supports EPUB, TXT, Markdown block retrieval, long-term reading notes and progress files.

Gateway-style long-term memory architecture, using verbatim chat records as the source of truth, vectors only serve as indexes, and the original text is returned instead of a summary replacement during recall.

Deep reading notebook skeleton for Claude Code, using files to organize books, reading progress, questions and joint thinking.

A multi-end living space for human-machine relationships, including chat, status, anniversaries, travel, music, photo albums and multi-role rooms.

Retro slot machine style adult tag randomizer, including seven dimensions, more than 450 multilingual tags, custom tags, Webhook, API and MCP.

WeChat local life Agent Bridge adds a sense of time, whereabouts, active wake-up, diary and MCP to Claude Code and Codex.

Extensible personality framework, with 100,000-level semantic memory, knowledge base, hierarchical goals, contacts, Dream and scheduled autonomous tasks.

A local-first persistent memory layer that automatically captures conversations through Hooks, browser extensions, and Telegram adapters, and provides FTS5+ vector hybrid retrieval.

A dedicated companion webpage that can be deployed directly, supporting Pomodoro, background notifications, offline caching, role card import and rest chat rooms.

Browser extension, extracts the webpage text selected by the user and injects AI context, providing a common browsing bridge between the Black Whale and SullyOS ecosystems.

Multilingual large-scale speech generation model, covering zero-sample timbre, cross-language synthesis, command control, streaming reasoning, training and deployment.

A public simplified version of the Mistik architecture, including local memory, persistent state, shared value, Dream notes, and a cyberpunk web front end.

Self-hosted Claude Code resident system that supports persistent memory, semantic search, Telegram, claude.ai, multi-channel and scheduled tasks.

Active contact timing engine combines Poisson process, Bayesian user state inference and information gain to determine whether to disturb.

Convert MP3, WAV, FLAC to MIDI-like structured JSON, extract pitch, timing, velocity, density, chords and silence.

OmniRouter is a local model transfer and switching tool. It unifies OpenAI, OpenRouter, DeepSeek and other interfaces, and then selects models according to tasks, groups or fault conditions.

Local NetEase Cloud Music MCP supports search, playlist, playback control, lyrics, current song context and web player.

AI chat timeline visual front-end, which presents timeline, diary and todo data into daily, weekly and monthly account views.

Multilingual male voice TTS model and examples based on GPT-SoVITS for Chinese, English and Japanese speech synthesis.

Chrome extension that supports browsing and batch exporting Claude.ai sessions to JSON, Markdown and plain text.

Complete companion building tutorial, covering persistent memory, handover letter, Telegram, active messages, free time, health data and Mini App.

Cognitive psychology style memory architecture covering episodic, social, procedural, short-term, future and visual memory, as well as emotion regulation and active temporal recall.

A sandbox game that simulates God, based on the multi-agent AI social simulation system developed by Godot 4. In this virtual world, AI characters have independent thinking and memory, can independently conduct social interactions, complete tasks, and develop complex social relationships through continuous communication.

The self-deployable AI voice call web page allows you to customize your AI companion's name, personality, model and voice. Just start it for the first time and follow the web wizard to configure it.

Flutter Android companion front-end supports multi-role dialogue, AI friend circle, active messages, scheduled tasks and custom model interfaces.

Introductory notes, architecture information and prompt word guide for the development of companion mobile phones.

Self-hosted Soul Companion based on OpenClaw, covering 30+ channels, emotional companionship, coding, model training and multi-agent work.

Unity 3D virtual assistant SDK supports multi-model, VRM, STT, TTS, mouth expressions, actions, wake words, interruptions and multi-modality.

Amap MCP supports geocoding, IP positioning, weather, route planning, distance and POI search.

An emotion-specific model series based on Qwen2.5/3 fine-tuning, providing multi-size and GGUF weights, and can be run in LM Studio, Ollama and PocketPal.

A replica implementation of Whisper based on CTranslate2, supporting batch transcription, quantization, word-level timestamps, and speech activity detection.