SpAItial opens partner beta for iPhone-captured multi-room 3D worlds

• SpAItial has opened a partner beta around iPhone capture: SpAItial_AI’s beta post promises consistent, high-quality 3D worlds from simple recordings, including multi-room scenes and generated coverage of missing gaps.
• Entry is application-based. In a reply, SpAItial_AI directs prospective creatives to its partner form and says the team will reach out.
• The company calls the system a generative world model in SpAItial_AI’s reply, while its earlier Echo-2 release notes describe a persistent 3D scene representation that can be explored in real time.
• A related creator experiment turned hundreds of 2D TED photos into a fly-through “spatial memory palace,” as bilawalsidhu’s TED post puts it.

A 54-second SpAItial_AI demo moves continuously between furnished interior rooms, then hands viewers an “Explore the 3D world” prompt. A separate k_grajeda clip depicts a foldout device and holographic interface, but its caption names no model, tool, or connection to SpAItial.

SpAItial’s beta post makes three concrete claims about the capture-to-world result:

1. A simple iPhone recording is the input.
2. One world can span multiple rooms.
3. Missing areas are filled while maintaining consistency and quality.

The announcement does not specify a recording path, supported iPhone models, capture duration, processing time, or world-size limit. One visible SpAItial_AI reply contains only an affirmative answer, adding no capture detail.

The company says the feature is in beta with partners and that its API is coming soon. Its reply points applicants to a creative-partner signup, without giving a timetable for selection or API availability.

An earlier SpAItial_AI partner-program announcement on LinkedIn says accepted applicants receive:

• Free credits
• Priority API access
• Early Echo model access
• Integration support
• A direct feedback loop with the lab

Asked directly whether the system is a generative world model, SpAItial_AI answered yes.

The company’s About page describes its spatial foundation models as systems that generate and reason about the appearance and physics of real and imagined 3D environments. Its homepage says the Echo family produces persistent 3D Gaussian Splat worlds for real-time exploration.

Bilawal Sidhu described a separate spatial capture built from “hundreds” of 2D TED photos, left where they were taken across the venue. The post does not identify the software used.

Its stated workflow has three parts:

• Lift 2D photos into 3D.
• Anchor the photos at their original locations.
• Fly through the venue and jump between the green room, backstage, and main stage.

When bilawalsidhu asked whether 360° input was supported, SpAItial_AI replied, “Yup!” The company’s product documentation separately lists text prompts, images, and 360° panoramas as inputs, then describes worlds as explorable, editable, shareable, and exportable.