Hextile Templates

Templates define how your 360° image is divided into hexagonal tiles for processing. Choosing the right template balances quality, processing time, and resource usage.

Template Overview

360 Hextile includes three preset templates:

Note: Times vary by pipeline — FLUX.1 Schnell processes in seconds per tile; Qwen-Image BF16 may take minutes per tile. Times shown are approximate for SDXL on RTX 4090.

Hextile_20_2K - Fast Processing

Specifications

• Total Tiles: 20 hexagons
• Tile Resolution: 2048 x 2048 pixels
• Coverage: Balanced distribution
• Processing Time: ~2 minutes (RTX 4090)
• VRAM Required: 6-8 GB

When to Use

• Quick previews: Test prompts and settings
• Iteration: Try multiple variations quickly
• Limited hardware: Works on entry-level GPUs
• Learning: Great for getting familiar with the system

Pros

• Fastest processing
• Lowest VRAM requirements
• Good for experimentation
• Quick turnaround

Cons

• Lower detail level
• May show slight seams
• Not suitable for final production
• Limited resolution scalability

Hextile_32_3K - Balanced (Recommended)

Specifications

• Total Tiles: 32 hexagons
• Tile Resolution: 3072 x 3072 pixels
• Coverage: Optimal distribution
• Processing Time: ~3 minutes (RTX 4090)
• VRAM Required: 8-10 GB

When to Use

• Production work: Best quality/time ratio ⭐
• General use: Covers most use cases
• Balanced approach: Good detail without excessive wait
• Mid-range hardware: Runs well on mainstream GPUs

Pros

• Excellent quality/speed balance
• Seamless results
• Handles fine details well
• Reasonable processing time

Cons

• Requires mid-range GPU
• Moderate processing time
• Higher than minimum VRAM

Recommended Settings

{
  "template": "Hextile_32_3K",
  "mode": "sequential",
  "strength": 0.4,
  "guidance_scale": 7.5,
  "steps": 35
}

Hextile_44_4K - Maximum Quality

Specifications

• Total Tiles: 44 hexagons
• Tile Resolution: 4096 x 4096 pixels
• Coverage: Maximum density
• Processing Time: ~4 minutes (RTX 4090)
• VRAM Required: 10-12 GB

When to Use

• Final production: Highest quality output
• Large prints: When every detail matters
• Professional work: Client deliverables
• High-end hardware: You have a powerful GPU

Pros

• Maximum detail preservation
• Perfect seam blending
• Handles complex textures
• Best for large outputs

Cons

• Longest processing time
• Requires high-end GPU
• High VRAM usage
• Overkill for web use

Recommended Settings

{
  "template": "Hextile_44_4K",
  "mode": "sequential",
  "strength": 0.35,
  "guidance_scale": 8.0,
  "steps": 40
}

Template Structure

Each template consists of:

Tile Geometry

Hexagonal tiles arranged in a geodesic pattern: - Even distribution across the sphere - Minimal distortion - Seamless overlap regions

Metadata Files

hextiles.json: Tile configuration

{
  "tiles": [
    {
      "tile_id": "tile_001",
      "position": [x, y, z],
      "rotation": [rx, ry, rz],
      "resolution": [4096, 4096]
    }
  ]
}

tilegen.json: Template parameters

{
  "num_tiles": 44,
  "tile_resolution": 4096,
  "overlap_percentage": 0.1,
  "projection_type": "equirectangular"
}

Inpaint Masks

Each tile includes an inpainting mask: - Defines overlap regions - Ensures seamless blending - Prevents edge artifacts

Choosing the Right Template

Decision Tree

Start
  ├─ Testing/Preview? → 20_2K
  ├─ Production Work?
  │   ├─ 6-8GB VRAM? → 32_3K
  │   └─ 10GB+ VRAM? → 44_4K
  └─ Final Deliverable? → 44_4K

By Use Case

Web Display - Resolution: Up to 4K - Template: 32_3K ✓ - Reasoning: Web viewers rarely show full detail

Print/Large Format - Resolution: 8K+ - Template: 44_4K ✓ - Reasoning: Print reveals all details

VR/Gaming - Resolution: 4K-6K - Template: 32_3K or 44_4K ✓ - Reasoning: Depends on target platform

Social Media - Resolution: 2K-4K - Template: 20_2K or 32_3K ✓ - Reasoning: Compressed for social platforms

Custom Templates

Create custom templates directly in the application using the Procedural Template Generator — no external tools required.

Custom Template Benefits

• Specific tile count: Match your needs exactly
• Custom resolution: Optimize for your hardware
• Specialized patterns: Adapt to specific content types
• Experimental: Try new approaches

Tile Ordering

The order in which tiles are processed affects the visual progression during rendering and can influence blending quality. Choose from these strategies in the Hextile tab:

Custom Ordering

Use the Copilot AI assistant to generate a custom tile order — describe your preferred sequence (e.g., "process the sky tiles first, then the horizon, then the ground") and the Copilot computes the optimal ordering.

Tile ordering is stored in the config as tile_ordering in the hextile section. Changing the ordering does not affect the final output — it only controls the processing sequence.

Procedural Template Generator

The built-in Procedural Template Generator lets you create fully custom hextile templates from 14 polyhedral geometries (icosahedron, dodecahedron, rhombicuboctahedron, and more). Key features include:

• 14 polyhedral base geometries with configurable subdivision levels
• Visual coverage map — see exactly how tiles distribute across the 360° sphere before generating
• Import/Export — share templates as ZIP archives or import community templates
• No Blender or external dependencies — everything runs inside the application

See Procedural Template Generator for the full guide.

Template Performance Comparison

Based on RTX 4090 (24GB VRAM) with SDXL pipeline. Times vary by pipeline — FLUX.1 Schnell processes in seconds per tile; Qwen-Image BF16 may take minutes per tile.

Processing Time by Steps

VRAM Usage by Mode

Tips and Best Practices

Start Small

Always test with 20_2K first: 1. Upload your image 2. Try different prompts 3. Find settings you like 4. Then move to 32_3K or 44_4K

Match Output to Use

Don't over-process: - Web? 32_3K is plenty - Instagram? 20_2K works fine - Print? Go with 44_4K - VR? Depends on platform

Consider Processing Time

Factor in your workflow: - Need quick iterations? → 20_2K - Overnight processing? → 44_4K - Balanced approach? → 32_3K

VRAM Headroom

Leave some VRAM free: - Don't max out your GPU - Keep 2GB free for system - Monitor temperature - Consider sequential mode

Troubleshooting Templates

Template Not Loading

Error: "Template not found"

Solutions: - Verify template folder exists - Check folder name matches exactly - Ensure all metadata files present - Re-download if corrupted

Out of Memory

Error: "CUDA out of memory"

Solutions: - Use smaller template - Switch to sequential mode - Close other applications - Reduce batch size

Poor Quality Results

Issue: Results look low quality

Check: - Using correct template for output size? - Sufficient diffusion steps? - Appropriate strength setting? - Input image quality adequate?

Next Steps

• Pipeline Settings - Configure AI pipelines
• Output Naming - Format and naming options
• Procedural Template Generator - Create custom templates
• Performance Tips - Optimize processing

Ready to process? Continue to Pipeline Settings →