360 Hextile Documentation

Welcome to 360 Hextile V2! Transform your 360-degree content with multiple AI pipelines, dynamic UI, and smart VRAM management.

🎉 What's New in V2

  • 3 AI Pipelines: SDXL Inpainting, Flux.2 Inpainting, Real-ESRGAN Upscaling
  • Dynamic UI: Backend defines forms, frontend renders automatically
  • Smart VRAM: Intelligent GPU memory management prevents crashes
  • Modular Design: Easy to add new AI models
  • Backward Compatible: All V1 workflows still work

Overview

360 Hextile V2 is a complete multi-pipeline processing system for 360-degree images. Choose from different AI models depending on your needs:

  1. SDXL Inpainting: Stable Diffusion XL for high-quality generation
  2. Flux.2 Inpainting: Advanced Flux model with faster convergence
  3. Real-ESRGAN Upscaler: AI-powered 4x upscaling (no tiling)

The system consists of:

  • Processing Pipelines: Modular AI processors with automatic registration
  • Hextile Templates: Hexagonal tile configurations for seamless 360° processing
  • Dynamic Frontend: Schema-driven UI that adapts to each pipeline
  • FastAPI Backend: Unified server with processor registry and VRAM management

🚀 Getting Started

New to 360 Hextile? Start here to learn the basics and get your system set up.

  • System requirements (GPU with 12GB+ VRAM recommended)
  • Installation guide
  • First render with SDXL pipeline

🎨 Pipeline Guide

Learn about the three available AI pipelines and when to use each.

  • SDXL Inpainting: Best for general-purpose 360° generation
  • Flux Inpainting: Advanced model with unique aesthetics
  • Real-ESRGAN: Fast 4x upscaling without prompts

🎬 Sequence Processing

Batch process image sequences for animation and time-lapses.

  • Temporal noise settings for smooth animations
  • Automatic frame management
  • Real-time progress tracking
  • Export completed frames in various formats

📦 Exporting Sequences

Export your rendered sequences in multiple formats.

  • Copy original files or convert formats
  • Resize to different resolutions
  • PNG/JPG conversion with quality control
  • Smart handling of partial sequences

📐 Templates

Learn about hexagonal tile templates and how to choose the right one.

  • 44_4K - High detail (44 tiles, 4K resolution)
  • 32_3K - Balanced (32 tiles, 3K resolution)
  • 20_2K - Fast processing (20 tiles, 2K resolution)

For Developers

🏗️ Architecture

Understand the V2 multi-pipeline system design.

  • Modular processor architecture
  • ProcessorRegistry and VRAM management
  • Dynamic UI schema system
  • Pipeline router and processing flow

🔧 API Reference

Complete FastAPI endpoint documentation for V2.

  • Processor endpoints (NEW)
  • Render job management with pipeline parameter
  • WebSocket real-time updates
  • LoRA model discovery (NEW)

🎨 Adding Pipelines

Create custom AI processing pipelines.

  • Extend BaseProcessor or TiledProcessor
  • Define UI schema for dynamic forms
  • Auto-registration system
  • VRAM management integration

Key Features

Multi-Pipeline Support (NEW in V2)

  • 3 AI Processors: SDXL, Flux.2, Real-ESRGAN
  • Easy Switching: Change pipelines with dropdown
  • Smart VRAM: Automatic model unloading prevents OOM
  • Extensible: Add new models easily

Dynamic UI System (NEW in V2)

  • Schema-Driven: Backend defines UI structure
  • Auto-Generated Forms: Frontend renders automatically
  • Conditional Fields: Show/hide based on selections
  • Special Components: Template picker, LoRA manager, Cardinal grid

AI-Powered Processing

  • Seamless 360° Tiling: Hexagonal projection system
  • Cardinal Prompts: Directional prompt interpolation
  • LoRA Support: Fine-tune with LoRA adapters
  • Flexible Templates: 20, 32, or 44 tile configurations

Modern Interface

  • Beautiful UI: Responsive dark mode interface
  • Real-Time Progress: WebSocket live updates
  • Tile Previews: See progress as tiles complete
  • Native Application: Built with Tauri
  • Config Save/Load: Preserve pipeline settings

Technology Stack

Backend

  • Python 3.10+ with FastAPI framework
  • ProcessorRegistry: Singleton pattern for pipeline management
  • CuPy: GPU-accelerated tile operations
  • PyTorch + Diffusers: AI model inference
  • Pydantic: Type-safe data validation

AI Models

  • Stable Diffusion XL: High-quality diffusion (~12GB VRAM)
  • Flux.2: Advanced diffusion model (~16GB VRAM)
  • Real-ESRGAN: AI upscaling (~2GB VRAM)
  • LoRA Support: Fine-tuning adapters

Frontend

  • SvelteKit: Reactive framework with TypeScript
  • shadcn-svelte: Beautiful UI components
  • TailwindCSS: Utility-first styling
  • Dynamic Forms: Schema-driven UI rendering

Desktop

  • Tauri 2.0: Rust-based native wrapper
  • WebView2: Modern web rendering engine

Need Help?

Quick Start

# Start the application
cd frontend
npm run tauri:dev

Launch the application and start creating!

Ready to create stunning 360° content with AI? Get started now →

Esc
Searching...
No results found.
Type to search the documentation
↑↓ Navigate ↵ Select Esc Close