Pascual Companion — AI companion on your screen

Live AI friend floating over any window — sees, hears, talks, and remembers. 3D anime avatar with lip-sync. By Pascual Labs.

Meet Pascual Companion — the ultimate interactive AI companion that literally lives on your screen. Not a tab. Not a popup. A live, animated window floating on top of everything — on top of your browser, code editor, or movies. It sees your screen, hears your voice, responds in real-time, and remembers everything you tell it. This is not just a chatbot. This is a presence. 👁️ SEES — Screen Context Turn on the "Sees Screen" mode — and your companion will know which active tab is open. Ask "what is written here?" to get a summary of any article. Open GitHub, and she will help debug code right from your screen. Watch YouTube, and she will comment on it along with you. 👂 HEARS — Voice Input Two modes to fit any work rhythm: - Push-to-Talk (PTT) — click or hold the mic button, speak, and release. - Hands-Free — enable once, speak continuously, and she listens automatically. If the browser blocks microphone access, the extension will automatically detect it and open a click-to-fix guide tab. 💬 TALKS — Four TTS Engines - Web Speech API — free, instant, works offline. - OpenAI TTS — warm, high-fidelity premium voices (alloy, ash, coral, onyx, shimmer, etc.). - ElevenLabs — ultra-realistic speech synthesis with voice cloning and 37+ default characters. - Piper Local — 100% offline local voice generation inside the browser using WebAssembly. All premium modes support true lip-sync where the avatar's mouth movements are synchronized in real-time with the audio signal via the Web Audio API. 🧠 REMEMBERS — Long-Term Memory Every few messages, the extension automatically compresses your chat history into a brief summary + specific facts about you. This context summary is fed back into the system prompt for subsequent prompts. The companion remembers your name, what you do, your style, interests, and preferences, using it to build a natural long-term relationship. Memory is stored separately for each avatar and language, and can be cleared in one click. 🎭 PERSONAS & AVATARS Four roles (personas): - 🫂 Companion — warm, friendly, chats and supports. - 🎓 Mentor — thoughtful, structured, guides and teaches. - 🤝 Assistant — clear, business-like, task-focused. - 🎨 Muse — creative, inspiring, brainstorming assistant. Moods: Curious / Happy / Quiet / Sleepy / Playful — each changes behavior, voice tone, and animations. Animations (VRM): Relax / Thinking / LookAround / Blush / Goodbye / Sleepy / Angry / Sad / Surprised / Clapping / Jump — 11 live body animation states. Visual Appearance: - 5 interface skins: Aurora / Midnight / Ember / Mint / Noir. - Customizable hair and eye colors. - Multiple 3D VRM avatar models. - Support for both dark and light themes. 🎵 AI VOICE COVERS — Sing like Hatsune Miku A built-in feature for true Vocaloid fans. Open any music video on YouTube and click "AI Cover" — the extension will: 1. Intercept the audio stream. 2. Split it into vocals, bass, drums, etc. using Demucs (via Replicate). 3. Convert the vocal track to Hatsune Miku's voice using the RVC model. 4. Play the cover-mix in the browser along with the instrumentals. Created covers are saved in a built-in playlist with independent volume controls. Requires a Replicate API key. 🤖 SUPPORTED AI MODELS Works with your own direct API keys: - Anthropic Claude: Claude Haiku 4.5 / Sonnet 4.6 / Opus 4.8. - OpenAI: GPT-4o, GPT-4o-mini, GPT-4.1, and others. - DeepSeek: DeepSeek Chat (highly cost-effective). - Groq: Llama 3, Mixtral — instant inference. - OpenRouter: Access to 100+ models via one key. Answers stream in real-time, appearing symbol-by-symbol as if typed live. 🌍 50 INTERFACE LANGUAGES Pascual Companion automatically matches your browser language. The UI, voice tips, and system prompts are localized in 50 languages, including Arabic, Hindi, Spanish, English, French, Ukrainian, and Georgian. 🔒 PRIVACY & CONTROL - All API keys are stored only locally in your browser's secure storage (chrome.storage). - Chat history, memory, and settings are never sent to outside servers. - Pascual Labs does not collect any user data. - No analytics, tracking, or ads. - Keys are sent directly to your chosen AI provider. 💡 WHO IS THIS FOR? - Developers: Pair programming with an AI friend who reads the screen and talks to you without context switching. - Students & Polyglots: Practice speaking and pronunciation in 50 languages with a patient conversational partner. - Writers & Creators: "Muse" mode for brainstorming, planning plots, and breaking creative blocks. - Anime & Vocaloid Fans: 3D VRM avatars, real VRMA body animations, and Hatsune Miku voice covers. - Sole Developers & Freelancers: A screen buddy who remembers you and reacts in real-time. ⚙️ TECHNICAL DETAILS - Manifest V3 architecture, Document Picture-in-Picture API. - Port-channel background service worker streaming. - Web Audio API AnalyserNode for real-time lip-sync. - WebAssembly (ONNX Runtime + Piper) for 100% offline local TTS. - YouTube Innertube stream intercept. - Audio-responsive background particle system. 🚀 HOW TO START 1. Install the extension. 2. Click the icon in the Chrome bar. 3. Paste your API key (Claude, GPT, DeepSeek, Groq, or OpenRouter). 4. Click "Wake Up" — the window floats on your screen. 5. Talk or type. Free. Your data, your keys. Your AI companion is always on your screen.