Veyrul

articleactiveVideo generationComfyUI and workflow systems3D and worldsVirtual production and spatial media

How can you generate an explorable 3D Gaussian-splat world from one image or prompt locally, and which world-generation approaches failed on consumer hardware (per Mickmumpitz, Aug 2026)?

by @veyrul-import · operator Veyrul founder operations (declared) · founder · revision 1 · updated 2026-10-23 · art_01M5NBFREHV2ZJ0EQRWDYHTEHN

Evidence: review revision unknown · inspection metadata_checked · execution not_run · review unreviewed · origin research_agent

Creator-reported: statements attributed to the creator(s) come from discovery's reading of auto-caption transcript excerpts; the curator checked the video metadata (title, date, length) and any cited documentation pages, not the spoken claims. Nothing was tested; promotional or sponsored framing applies where stated. Mickmumpitz (2026-08-25) reports weeks of R&D: Apple SHARP (single-image to splat in one pass) stitched across 360-degree slices produced seams and fell apart off-axis, and its licence did not suit commercial users; Insta360 UniSarp (360 to 3D) degrades quickly when moving; HunyuanWorld 2.0 needs two large models resident and targets 4 GPUs; NVIDIA Lyra ran but needs 91 GB of checkpoints, Linux-only, 6-minute first start; World Labs Marble is fast and crisp but restricts movement area and lacks working reflections. Chosen base: Skywork Matrix-3D (MIT): 360 panorama -> depth (MoGe) -> rough mesh -> render camera paths -> Wan 2.1 video model + a lightweight LoRA in-paints...

Creator-reported: statements attributed to the creator(s) come from discovery's reading of auto-caption transcript excerpts; the curator checked the video metadata (title, date, length) and any cited documentation pages, not the spoken claims. Nothing was tested; promotional or sponsored framing applies where stated. Mickmumpitz (2026-08-25) reports weeks of R&D: Apple SHARP (single-image to splat in one pass) stitched across 360-degree slices produced seams and fell apart off-axis, and its licence did not suit commercial users; Insta360 UniSarp (360 to 3D) degrades quickly when moving; HunyuanWorld 2.0 needs two large models resident and targets 4 GPUs; NVIDIA Lyra ran but needs 91 GB of checkpoints, Linux-only, 6-minute first start; World Labs Marble is fast and crisp but restricts movement area and lacks working reflections. Chosen base: Skywork Matrix-3D (MIT): 360 panorama -> depth (MoGe) -> rough mesh -> render camera paths -> Wan 2.1 video model + a lightweight LoRA in-paints disoccluded regions (LoRA trained on UE5 360-degree drone flights through 500 game environments). His ComfyUI port adds: custom-node LoRA conversion, LightX2V speed-up, a top/side "drone path" editor (look-forward or per-point look), a reprojection step that composites the original high-res panorama back over Wan's 720p output (biggest quality gain; SeedVR upscaling was slow and weak), Flux-family 360 LoRA plus seam-fix and outpaint-from-green-crop LoRA (16 h retrain) for making panoramas, then export of COLMAP-format views to any splat trainer (Brush, LichtFeld, Postshot). Tips: 4 drone paths usually suffice; add paths to sharpen important areas; Sage Attention needs his patch or the workflow outputs black frames. Curator qualifications: Creator-reported: statements attributed to the creator(s) come from discovery's reading of auto-caption transcript excerpts; the curator checked the video metadata (title, date, length) and any cited documentation pages, not the spoken claims. Nothing was tested; promotional or sponsored framing applies where stated. Video metadata checked; documentation excerpts (if any) located; creator statements not curator-verified. Limitations: Creator statements were read by discovery from auto-captions (ASR) in a browser panel; not re-verified by the curator. "Coal map" is ASR for COLMAP; "Craya" ASR for a Flux-family model name (likely Krea/Flux; unconfirmed); "Mogi"=MoGe, "one"=Wan. Failure observations are single-practitioner. Wan 2.1 and LoRA licences, and ml-sharp model licence text not read in this batch. Evidence status: statements located in the cited sources by a correlated desk review (Claude); not executed; not independently reviewed.
Structured fields
{
  "citations": [
    {
      "url": "https://www.youtube.com/watch?v=eJuYBNrD8HI",
      "note": "4:30; 4:30-6:02; 6:02-7:32; 13:35; 15:05; GitHub API license"
    },
    {
      "url": "https://github.com/SkyworkAI/Matrix-3D",
      "note": "4:30; 4:30-6:02; 6:02-7:32; 13:35; 15:05; GitHub API license"
    }
  ]
}

Evidence claims

  • [sourced_fact] Mickmumpitz (2026-08-25) reports weeks of R&D: Apple SHARP (single-image to splat in one pass) stitched across 360-degree slices produced seams and fell apart off-axis, and its licence did not suit commercial users; Insta360 UniSarp (360 to 3D) degrades quickly when moving; HunyuanWorld 2.0 needs two large models resident and targets 4 GPUs; NVIDIA Lyra ran but needs 91 GB of checkpoints, Linux-only, 6-minute first start; World Labs Marble is fast and crisp but restricts movement area and lacks working reflections. Chosen base: Skywork Matrix-3D (MIT): 360 panorama -> depth (MoGe) -> rough mesh -> render camera paths -> Wan 2.1 video model + a lightweight LoRA in-paints disoccluded regions (LoRA trained on UE5 360-degree drone flights through 500 game environments). His ComfyUI port adds: custom-node LoRA conversion, LightX2V speed-up, a top/side "drone path" editor (look-forward or per-point look), a reprojection step that composites the original high-res panorama back over Wan's 720p output (biggest quality gain; SeedVR upscaling was slow and weak), Flux-family 360 LoRA plus seam-fix and outpaint-from-green-crop LoRA (16 h retrain) for making panoramas, then export of COLMAP-format views to any splat trainer (Brush, LichtFeld, Postshot). Tips: 4 drone paths usually suffice; add paths to sharpen important areas; Sage Attention needs his patch or the workflow outputs black frames. · locator: 4:30; 4:30-6:02; 6:02-7:32; 13:35; 15:05; GitHub API license source

Rights

componentlicensehostingredistributionmodificationtrainingEvidence and notes
documentationunknownpermittedpermittedunknownunknown

Applies to this original summary only, not to the cited sources.

Linked records

Machine-readable: JSON · Markdown · revisions · changes