User Interface Guide

The 360 Hextile application provides an intuitive, modern way to process your 360° images with real-time feedback.

Overview

The application is built with SvelteKit and Tauri, offering:

• Drag-and-drop upload: Easy image uploading
• Real-time progress: Live tile previews as they process
• Job management: Track multiple processing jobs
• Dark mode: Comfortable viewing in any lighting
• Native performance: Direct file system access

Getting Started

Launching the Application

Development Mode:

cd frontend
npm run tauri:dev

Production Build:

cd frontend
npm run tauri:build
# Find installer in src-tauri/target/release/bundle/

This automatically: 1. Starts the backend API server 2. Launches the native application window 3. Connects to the processing backend

Interface Overview

Main Navigation

The application has two main pages accessible from the navigation menu:

• Home: Main render configuration interface
• Pipelines: Pipeline management (install, repair, uninstall AI models)
• Settings: Application preferences and authentication

Configuration Tabs

The home page is organized into tabs that vary based on the selected pipeline:

Standard Tabs (all pipelines): 1. Input: Upload and configure your 360° image 2. Hextile: Select template and tile settings 3. Prompts: Global and cardinal direction prompts 4. Generation: AI processing parameters (strength, steps, guidance) 5. LoRA: Select and configure LoRA models 6. Output: Set output format and quality

Pipeline-Specific Tabs: - ControlNet: Configure structural guidance (SD 1.5, SD 2.1 only) - Temporal: Propagation settings for sequences (SDXL only)

Workflow

Select Pipeline → Upload Image → Select Template → Configure AI → Start Processing → Download Result

Tab 1: Input

Uploading Images

Drag and Drop: 1. Drag your 360° image onto the upload area 2. Image preview appears automatically 3. File information displays (size, dimensions)

Click to Browse: 1. Click the upload area 2. Select file from file browser 3. Supported formats: JPG, PNG

Requirements: - Equirectangular projection (2:1 aspect ratio) - Minimum 2048x1024 pixels - Maximum 100MB file size

Image Preview

Once uploaded, you'll see: - Thumbnail preview - File name and size - Dimensions (width x height) - Remove button to clear and upload new image

Input Configuration

The Input tab provides three key settings in a responsive grid:

Resolution Settings (Width & Height) - Configure input image dimensions - Default: 8192 x 4096 (8K equirectangular) - These values should match your source image dimensions

Playback FPS - Sets the default playback speed for sequence animations - Options: 5, 10, 15, 24 (film standard), 30 (default), 60, 120 FPS - Only applies to sequence mode - Can be changed during playback on the sequence details page (including custom FPS values) - Saved to configuration file for future renders

Tab 2: Hextile Templates

Template Selection

Choose from three preset templates:

Hextile_20_2K - Fast - 20 hexagonal tiles - 2K resolution per tile - ~2 minutes processing (RTX 4090) - Good for previews and testing

Hextile_32_3K - Balanced ⭐ Recommended - 32 hexagonal tiles - 3K resolution per tile - ~3 minutes processing (RTX 4090) - Best quality/speed ratio

Hextile_44_4K - Maximum Quality - 44 hexagonal tiles - 4K resolution per tile - ~4 minutes processing (RTX 4090) - Highest detail output

Processing Mode

Sequential Mode (Default) - Processes tiles one at a time - Lower VRAM usage - More predictable results - Recommended for most users

Batch Mode - Processes multiple tiles simultaneously - Faster if you have high VRAM - Requires 12GB+ VRAM - May produce slight variations

Tab 3: Diffusion Settings

Prompt

Describe the style or modifications you want:

Examples: - "vibrant colors, enhanced saturation" - "photorealistic, ultra detailed" - "oil painting style" - "cyberpunk aesthetic, neon lights"

Tips: - Be specific and descriptive - Use positive terms (what you want) - Separate concepts with commas - Reference art styles or moods

Strength (0.0 - 1.0)

Controls how much AI modifies the original:

• 0.0 - 0.3: Subtle enhancement, keeps original mostly intact
• 0.4 - 0.6: Moderate changes, noticeable style application
• 0.7 - 1.0: Heavy modification, may deviate significantly

Recommended: Start with 0.3-0.5

Guidance Scale (1 - 20)

How closely the AI follows your prompt:

• 1 - 5: Loose interpretation, more creative
• 6 - 10: Balanced adherence to prompt
• 11 - 20: Strict adherence, less creativity

Recommended: 7-10

Steps (10 - 100)

Number of diffusion iterations:

• 10 - 20: Fast, lower quality
• 20 - 50: Good balance (recommended)
• 50 - 100: Highest quality, very slow

Recommended: 30-40

Seed

Random seed for reproducibility:

• -1: Random seed (different each time)
• Any number: Fixed seed (reproducible results)

Use a fixed seed to reproduce exact results.

Temporal Noise (Sequence Mode Only)

When processing image sequences, additional noise strategy options appear to prevent static noise patterns in animations:

• Tile Increment: Original behavior (may show static patterns)
• Frame Offset: Recommended - shifts noise each frame
• Frame Hash: Hash-based variation for more randomness
• SLERP: Smooth interpolation between noise keyframes (~30ms overhead)
• Optical Flow: Motion-guided noise for highest quality (~300ms overhead)

See Sequence Processing Guide for detailed information.

ControlNet Tab (SD 1.5, SD 2.1 only)

When using SD 1.5 or SD 2.1 pipelines, the ControlNet tab appears for configuring structural guidance:

Enable ControlNet

Toggle to activate ControlNet processing. When enabled, additional options appear.

ControlNet Type

Select the type of structural guidance:

SD 1.5 Options: - Canny, Depth, OpenPose, Scribble, MLSD, Seg, Normal, Lineart

SD 2.1 Options: - Canny, Depth, OpenPose, Scribble

Strength & Timing

• Strength: How strongly ControlNet influences generation (0.0-2.0)
• Guidance Start: When ControlNet begins (0.0-1.0, fraction of steps)
• Guidance End: When ControlNet ends (0.0-1.0, fraction of steps)

Preprocessor Settings

Type-specific settings like Canny thresholds or depth boost appear based on selection.

See ControlNet Guide for detailed information.

Temporal Tab (Sequences only)

When processing sequences with SDXL pipeline, the Temporal tab provides advanced animation settings:

Temporal Propagation

Enable keyframe-based animation that only runs full diffusion on every Nth frame:

• Keyframe Interval: Frames between full diffusion (2-48)
• Keyframe Strength: Diffusion strength for keyframes
• Propagation Mode: Latent space (smoother) or Pixel space (sharper)
• Motion Source: Auto, EXR only, Optical Flow, or None

Blending & Repair

• Bidirectional Blend: Blend from both previous and next keyframes
• Blend Falloff: Cosine (smoother) or Linear
• Repair Settings: Configure threshold, strength, and ControlNet for fixing low-confidence areas

See Sequence Processing Guide for detailed information.

Output Tab

Resolution

Choose output resolution: - Match Input: Same as input image - 2K: 2048x1024 - 4K: 4096x2048 - 8K: 8192x4096 (requires high VRAM)

Format

• PNG: Lossless, larger files
• JPG: Compressed, smaller files

Quality (JPG only)

• Low (50-70): Small files, visible artifacts
• Medium (70-85): Balanced
• High (85-95): Large files, minimal artifacts
• Maximum (95-100): Very large, imperceptible loss

Jobs Tab

Job List

View all your processing jobs:

• Status: pending, running, completed, failed
• Progress: Percentage complete
• Time: Started, duration
• Actions: View, Download, Delete

Job Details

Click any job to see:

• Input image preview
• Current processing tile
• Progress bar
• Real-time log messages
• Tile preview (updates as processed)

Real-Time Updates

Watch your job process in real-time:

• Progress bar: Shows completion percentage
• Current tile: Which tile is processing
• Tile preview: See processed tiles appear
• Status messages: Detailed processing log

Downloading Results

Once complete: 1. Click "Download" button 2. Save to your preferred location 3. View in any 360° image viewer

Pipeline Management Page

Access via Pipelines in the navigation menu. This page lets you manage AI model installations.

Pipeline List

Each pipeline card shows: - Name: Pipeline display name (e.g., "Stable Diffusion 3.5") - Status: Installed, Not Installed, Installing, Requires Auth - Description: Brief description of the pipeline - Actions: Install, Repair, Uninstall, Validate, Test

Installing Pipelines

1. Find the pipeline you want to use
2. Click Install
3. For gated models (SD 3.5, Flux), you'll be prompted to authenticate
4. Monitor download progress via the progress bar
5. Once complete, the pipeline becomes available for use

HuggingFace Authentication

Some models require HuggingFace authentication: 1. Click Authenticate with HuggingFace when prompted 2. Enter your access token (get one from huggingface.co/settings/tokens) 3. Click Validate 4. Once authenticated, gated model downloads will proceed

Pipeline Actions

See Pipeline Guide for detailed information.

Settings Page

Access via Settings in the navigation menu (gear icon).

GPU Information

View your system's GPU status: - GPU model name - VRAM total and used - CUDA and driver versions

This helps determine which pipelines are suitable for your hardware.

Output Folder

Configure where renders are saved: - Browse to select a custom output directory - Default: output/ in the application directory

Render Options

Toggle application behaviors: - Auto-cleanup tiles: Delete intermediate tile files after render - Auto-open folder: Open output folder when render completes - Save config with renders: Save .hextile.json config alongside output

HuggingFace Authentication

Manage your HuggingFace token for gated model access: - Enter or update your access token - Validate token status - View which models require authentication

See Settings Guide for detailed information.

Tips and Tricks

Best Practices

1. Start small: Test with 20_2K template first
2. Iterate: Try different prompts and settings
3. Save configs: Note settings that work well
4. Monitor VRAM: Check GPU usage during processing

Performance Optimization

• Use Sequential mode for limited VRAM
• Close other GPU applications
• Process during off-peak times
• Use lower resolution for testing

Troubleshooting

Upload fails: - Check file format (JPG/PNG only) - Verify image is equirectangular - Ensure file size < 100MB

Processing stalls: - Check API server is running - Verify GPU has enough VRAM - Look for error messages in job log

Slow processing: - Choose smaller template (20_2K) - Reduce steps to 20-30 - Close other applications - Check GPU temperature

Keyboard Shortcuts

• Ctrl/Cmd + U: Upload image
• Ctrl/Cmd + S: Start processing
• Tab: Navigate between tabs
• Esc: Close modals

Next Steps

• Pipeline Guide - Detailed pipeline documentation
• ControlNet Guide - Using ControlNet for structural guidance
• Settings Guide - Application preferences
• Sequence Processing - Animation and temporal features
• Templates - Learn about template creation
• Troubleshooting - Fix common issues

Ready to start? Check out the Getting Started Guide for a step-by-step walkthrough.