Format Converter
Convert between 360° image projection formats
Last updated: 2026-03-13
Format Converter
Overview
Convert between 360° image projection formats without AI processing. GPU-accelerated mathematical remapping — instant results, no model download required.
Supported Formats
Input Formats
- Equirectangular — Standard 2:1 360° format (most common)
- Fulldome (Domemaster) — Circular fisheye for planetarium domes
- Fulldome Angular — Angular fisheye variant
- Cubemap Cross — 6-face cube unfolded into a cross layout
Output Formats
- Equirectangular — Universal 360° interchange format
- Fulldome (Domemaster) — For planetarium projection
- Fulldome Angular — Angular fisheye variant
- Cubemap Cross — For game engines, VR applications
- Rectilinear — Standard flat/perspective view extraction
Conversion Types
Equirectangular ↔ Cubemap
Convert between the standard 360° format and 6-face cubemap layout. Useful for game engine integration (Unity, Unreal, Godot).
Equirectangular ↔ Fulldome
Convert for planetarium dome projection. Supports both standard domemaster and angular fisheye variants.
Rectilinear Extraction
Extract a flat (perspective) view from any direction in a 360° image: - Set field of view angle (how wide the crop) - Set look direction (yaw + pitch) - Output is a standard flat image — use for social media, prints, previews
Usage
The Format Converter is a processor selected in the Diffusion tab's pipeline dropdown — not a separate tab. Select format_converter from the pipeline list.
Single Image
- In the Diffusion tab, choose Format Converter from the pipeline dropdown
- In the Input tab, upload or select your source image
- In the Output tab, choose your desired output format
- Start processing — the conversion runs instantly
Batch Conversion
Use the batch convert component to process multiple images at once: 1. Select the Format Converter pipeline 2. Switch to batch mode in the Input tab 3. Add multiple images to the batch queue 4. Configure output format and destination 5. Start batch processing — all images convert in sequence
Technical Notes
- All conversions route through equirectangular as interchange format
- GPU-accelerated via CuPy CUDA kernels
- Professional-grade quality at any resolution up to 8K
- No AI model required — pure mathematical remapping
- 0 GB VRAM required (no AI model loaded)