Skip to content

ROADMAP.md

Latest commit

 

History

History
53 lines (41 loc) · 2.91 KB

ROADMAP.md

File metadata and controls

53 lines (41 loc) · 2.91 KB
Raw

Roadmap

What is being worked on next, roughly in order of priority. This is a direction, not a set of commitments or dates.

Ideas and requests are welcome as issues.

Next

  • Migrate the VR viewer to the current splat renderer. The main viewer moved to Spark; the WebXR viewer still uses the previous library, which is why both are still dependencies. Finishing this removes one, and gives VR support for the SOG format for free.
  • Verify VR against the current Three.js. The engine moved to three 0.180 for the renderer migration; the VR path has not been re-tested in a headset since.
  • Exercise the remaining splat formats. .ply splats and .ksplat are supported and mapped but no bundled story uses either, so nothing confirms those loader paths against the current renderer.

Authoring

  • Warn about missing assets before export, rather than at import time. A story that references a file the app never received currently exports quietly and fails for the reader.
  • Undo/redo in the editor.
  • Multiple models in one story, so a story can move between locations rather than being anchored to a single scan.
  • Point-cloud polish — percentile-based framing for point clouds (splats already have it), a warning for very large clouds, and a per-file up-axis override for scans that are not Z-up.

Reading

  • Accessibility audit. Captions, transcripts, alt text, keyboard navigation and screen-reader behaviour across both modes. Currently unaudited, and the immersive mode in particular needs a considered non-visual story.
  • Mobile refinement of the immersive view — the overlay is workable but was designed on desktop.
  • Transition polish between sections and between the two modes.
  • Travel between stories. A reader currently has to return to the gallery to move on. Add a next/previous path (ordered by the gallery) and optional curated links between stories, so a set can be read as a sequence or a web. The scene swap should be seamless — keeping the 3D context alive rather than reloading — with controls in both reading modes. Targets come from the deployment's own story index, so a single-story export simply shows none.

Platform

  • Deduplicate shared scans. Today each exported story bundles its own copy of the model, so two stories about one building carry it twice. Splitting the story from its scan is the first real step from a single-author engine toward the shared-infrastructure idea in the vision.
  • Story discovery — a map or index browse mode, secondary to stories rather than replacing them.

Anything requiring accounts, server-side storage or moderation is out of scope for the engine by design. Those belong to a platform layer built on top of it — see Vision → The engine and the platform.