LLM Hub - Local AI Assistant

Advanced AI runs completely on your phone - no internet required, no data shared, complete privacy guaranteed. State-of-the-art Large Language Models in your pocket! TEN POWERFUL AI TOOLS SMART CHAT: Multi-turn chats with context awareness, RAG memory, optional web search, TTS, and text/image/audio inputs. Create custom AI personas. WRITING AID: Summarize docs, expand ideas, rewrite text, fix grammar, and generate code. IMAGE GENERATOR: Create images with Stable Diffusion 1.5. Generate variations and upscale resolution on-device. TRANSLATOR: Real-time offline text, image (OCR), and audio translation for 50+ languages. TRANSCRIBER: Convert audio to text offline with high accuracy using Whisper ASR models. VIBECODE Creator: Build custom coded AI personas to control AI behavior across chat sessions. CODE ASSISTANT: Generate, debug, and explain code with optional web search for docs. SCAM DETECTOR: Analyze messages and images for phishing attempts with detailed risk explanations. IMAGE UPSCALER: Enhance and upscale image resolution and details on-device. VIBEVOICE: Hands-free real-time voice chat with local AI. CUTTING-EDGE AI MODELS Formats: GGUF, ONNX, LiteRT, MediaPipe (MediaPipe, ONNX Runtime, Nexa SDK) • Gemma-3 1B (Google) - Fast and efficient • Gemma-3n E2B/E4B (Google) - Multimodal (text, vision, audio) • Gemma-4 E2B/E4B (Google) - Multimodal (text, vision, audio) • Gemma-4 12B (Google) • Llama-3.2 1B/3B (Meta) - Powerful open-source • Phi-4 Mini (Microsoft) - Mobile-optimized • Whisper ASR Family - Offline speech-to-text • Ministral 3B/8B (Mistral AI) - Multilingual with vision • GPT-OSS Family - Fast ONNX models • LFM-2.5 Thinking - Reasoning and long-form • Granite Models (IBM) - Enterprise-ready • Stable Diffusion 1.5 - On-device image gen and upscaling PRIVACY & SECURITY • Zero data collection - everything stays on device • No internet required for AI inference • No accounts, tracking, or cloud uploads • Open-source codebase ADVANCED FEATURES • Text-to-Speech with auto-readout • GPU/NPU acceleration for fast performance • Multimodal support (text, images, audio) • RAG with global memory • Custom AI personas (creAItor) • On-device image upscale and detail enhancement • Custom model import (.gguf, .onnx, .task, .litertlm, .mnn) • Direct downloads from HuggingFace • MediaPipe, ONNX Runtime, and Nexa SDK backends • Beautiful Material Design UI with 17 languages HOW IT WORKS 1. Download models in-app (one-time) 2. Create custom AI personas with VibeCode 3. Choose your tool (Chat, Writing, Image Gen, Translate, Transcribe, Code, Scam Detector, Image Upscaler, or VibeVoice) 4. Use completely offline with full privacy PERFECT FOR • Privacy-conscious users • Professionals needing offline AI • Developers coding and debugging • Artists generating AI images • Students and researchers • Travelers needing offline translation • Protecting against phishing scams OPEN SOURCE github.com/timmyy123/LLM-Hub