Synthetic World Engine · Live

w0rldw3aver360

CC0 PANORAMA DATASET → LoRA → GAUSSIAN SPLATS

Not a screenshot gallery. Drag inside any sphere — this is the actual training data, rendered equirectangular and explorable. 12 curated worlds, 6 NASA public-domain 360 video clips, every caption engineered to teach a diffusion model the projection format itself.

12
Still Worlds
6
360 Video Clips
775
Caption Words
CC0
License
2:1
Equirect
Loading…
DRAG look around
SCROLL zoom
0.0° / 0.0°
Loading sphere

Training Caption

01

Why this dataset hits different

Most 360 datasets are YouTube rips or gated robot rosbags with no license. This one is clean, captioned in a uniform grammar, and built so the format is learnable — not just the subject.

🧬

Uniform caption DNA

Every caption opens with the trigger token, then the identical scaffold: w0rldw3aver360, a 360-degree equirectangular photograph of…. 56–72 words, 12/12 files. The repetition is the point.

🌀

Format, not just subject

Captions describe the wrap: horizon circling the frame, seamless seam, zenith and nadir. Rendered cold, the base model produced true 2:1 equirect from caption alone — scored 9/10.

⚖️

License-clean

Every still is CC0 from Poly Haven. Every video is US Government public domain (NASA/JPL/MSFC). Curation and captions released CC0. No gray area to litigate later.

🎯

Pairs, drag-and-drop

A pairs/ folder ships in the repo — .jpg + identically-named .txt side by side. Drag the whole folder into AI-Toolkit and train. No renaming script needed.

🧊

Built for splats

15-second 2:1 clips at 30fps are the right shape for Gaussian splat training — enough parallax to reconstruct geometry, short enough to stay coherent.

🔁

Synthetic expansion

The real pipeline: train the LoRA on these 12, then generate hundreds of new worlds with Wan I2V and feed those back in. The dataset bootstraps itself.

02

Motion — 360 video

NASA public-domain equirectangular video, cut to 15-second windows and cropped to exact 2:1. Flat renders and partial mosaics were QC-rejected.

Clips
03

Narrative test — "Meeting Sasquatch"

Proof that chained generation holds a story together. Five MiniMax H3 clips, each starting from the previous clip's final frame. One still panorama → a 26-second first-person short film with synchronized audio.

01 · The Walk
Cautious POV through old growth
02 · The Encounter
It steps out from the trees
03 · The Offer
A cupped palm, glowing mushrooms
04 · The Trip
The forest breathes and blooms
05 · The Departure
It fades back into the trees
Seam integrity (PSNR between clips) 38.438.140.834.8 dB — above 30 is visually seamless
04

The pipeline

End to end, this is what the studio runs. Each stage is a real artifact that exists right now.

Source & filter

Poly Haven CC0 stills; NASA 360 video pulled from the public images API. Rejected: Curiosity 8K (partial Navcam mosaic with burned-in text) and Hurricane Maria (flat data-viz, not a panorama).

Curate to spec

ffprobe every candidate for true 2:1 equirect. Crop letterboxed sources to exact ratio — Perseverance needed crop=1752:876:72:40 to kill the zenith black bar.

Caption with intent

Trigger token first, uniform scaffold, 56–72 words describing subject, horizon, lighting, and wrap behavior. Captions are CC0 too.

Package

Published to HuggingFace with pairs/ ready for drag-and-drop, README with YAML front-matter, and a montage preview.

Train

AI-Toolkit config targeting Krea 2 Turbo: rank 32, 1024×512 for panos, 3000 steps, low_vram + cache_text_embeddings for 12GB cards. No trigger override — captions carry it.

Generate & feed back

Render new worlds with the trained LoRA, motionize with Wan I2V, feed the best back into the dataset. Synthetic data flywheel.

The pipeline is the product

Datasets, LoRAs, sites, and releases — built by an agent swarm on a home GPU and a $20 runway. This page rendered itself from the same repo it documents.

Copied