Overview: February develops a browser-based VR sandbox for translating generative visual material into immersive environments. Building on the publication of Oblatio and initial Gaussian splat experiments, this phase focuses on integrating AI-generated video, splat-based scene construction, and a custom three.js player into a unified workflow. The aim is to establish a repeatable pipeline for producing and delivering 360° and fulldome immersive experiences derived from symbolic source material.1
Month: Febuary 2026
Media: VR, generative video, 360 degree video, environment composition
PhD Stage: Year 2
Key Themes: Awe, liminality, ritual aesthetics, spatial presence
Category: Research Practice
Research Question: How can immersive environments be composed to support awe and transformative reflection?
Project Strand: 360 degree stereoscopic spacial sound VR prototyping
Output: Three.js platform prototype development and reflective documentation
Practice
The focus of this phase is the construction of an operational workflow that moves from symbolic image production into immersive delivery. Using paintings and generative compositions as source material, environments are developed through two parallel processes: AI-generated video sequences and Gaussian splat reconstructions. These are then forced into 360° equirectangular formats using tools such as Google Flow, enabling their deployment within immersive viewing contexts.
To support this, a custom three.js player has been developed as a lightweight delivery infrastructure for browser-based immersive media. The system enables playback of 360° video environments, supports scene transitions, and introduces basic interaction logic for navigating between symbolic environments.
This phase is not concerned with producing finished works, but with establishing a coherent pipeline in which generative imagery can be stabilised, spatialised, and presented as an immersive environment. The underlying question is whether symbolic and generative media can be organised into a system capable of supporting repeatable immersive experience design.
Experiments and Iterations
The first stage involves the use of paintings and collage-based compositions as substrates. These are processed through AI video systems to produce animated sequences that retain aspects of the original symbolic structure while introducing motion, atmosphere, and temporal transformation. Parallel to this, selected sequences are used as inputs for Gaussian splat generation, producing volumetric approximations of environments derived from the same source material.
Contemporary generative video systems are not designed to produce spherical or equirectangular outputs. However, through iterative prompting, a set of conditions was identified under which the system produced imagery exhibiting sufficient radial coherence to be interpreted as 360° environments.
The three.js player functions as the integration layer. It maps equirectangular video onto an inverted sphere, enabling full 360° viewing within the browser. Iterations focused on interaction design, including field-of-view manipulation (in place of positional camera movement), crossfading between scenes, and minimal UI overlays to avoid breaking immersion. An idle dimming system was introduced to modulate attention and reduce interface dominance.
These experiments are exploratory and often produce inconsistent outputs. However, they establish a working pipeline linking symbolic source material, generative transformation, volumetric reconstruction, and immersive delivery.
Artefacts and Documentation
Generative Video Environment Set
A series of AI-generated video sequences derived from paintings and compositional studies. These sequences explore motion, atmosphere, and transformation within symbolic environments, and serve as primary inputs for both direct 360° conversion and splat-based reconstruction. They form the temporal layer of the immersive pipeline.
Gaussian Splat Scene Experiments
Early-stage volumetric reconstructions generated from AI imagery and video frames. These splat-based environments provide a spatial counterpart to the video sequences, suggesting a method for stabilising generative outputs into reusable scene structures. While technically unstable, they demonstrate the potential for persistent, navigable environments derived from generative sources.
Technical and Process Notes
Three.js VR Player Prototype
A custom browser-based immersive media player was developed to support the delivery of 360° equirectangular video environments for VR-style and fulldome-oriented presentation. The system allows playback of panoramic scenes from a fixed central viewpoint, with rotational viewing, adjustable field of view, scene switching, crossfade transitions, and a minimal interface that dims during inactivity. These features were designed to preserve atmosphere and symbolic continuity while providing a lightweight testing ground for generative immersive environments. The player functions as a delivery layer within the wider workflow, translating generated and symbolic media into navigable experiential form.
Several design choices were made deliberately:
- Fixed-centre viewing preserves coherence in 360° material
- Crossfades maintain atmosphere between scenes
- Minimal UI protects immersion
- FOV modulation supports the perception of scale and vastness
- Browser deployment allows rapid iteration and accessibility
Critical Reflection
For the first time, I was able to encounter my own generated material in a frameless, spherical, all-enclosing form. This changed the work from an idea about immersion into an iterative confrontation with it. The question was no longer simply how to produce symbolic or generative imagery, but whether such imagery could become an environment with enough scale, coherence, and atmospheric force to sustain attention, invite absorption, and begin to approach awe.
This has sharpened the research considerably. The central problem is not delivery infrastructure in itself, but how emerging tools can be used to create worlds that one would actually want to remain inside: environments with sufficient fidelity, spatial plausibility, pacing, and symbolic depth to support contemplative or intensified states rather than merely producing novelty. In methodological terms, this confirms the practice-based basis of the doctorate. Such questions cannot be settled abstractly. They have to be worked through iteratively, by testing thresholds of scenic sequence, perceptual scale, world-coherence, and atmospheric charge.
This phase clarified the importance of infrastructure within the practice. While previous work focused on symbolic composition and generative image production, the development of the three.js player foregrounded the necessity of a delivery system capable of organising and presenting these materials coherently.
The experiments suggest a possible transition from generative imagery as transient output toward immersive environments as structured experiences. However, this remains unresolved. The workflow is unstable, and the outputs often lack consistency in visual quality and spatial coherence. There is a risk that the process produces technically novel results without sustaining symbolic or aesthetic depth.
What has emerged is a clearer articulation of the problem: not how to generate compelling images, but how to construct environments that can persist, be revisited, and support meaningful engagement. The question of persistence—technical, aesthetic, and symbolic—now appears central to the development of the work.
Research Development
Following Grau (2003), immersive environments can be understood as part of a historical continuum of image-world construction, while Turner’s concept of liminality (1969) frames these environments as threshold spaces in which symbolic transformation occurs. The current workflow translates these principles into a technical pipeline, in which traditonally created symbolic material is progressively transformed into immersive environments.
The experiments engage with contemporary understandings of awe as arising from perceived vastness and the need for accommodation (Keltner and Haidt 2003), further elaborated in immersive contexts as epistemic expansion and emotional complexity (Chirico et al. 2022). By constructing 360° environments and volumetric approximations, the work explores how such conditions may be produced and modulated through spatial media.
Drawing on Latour (1999) and Star (1999), the development of the three.js player reframes immersive media as infrastructural rather than purely expressive. The system functions as a platform for organising and delivering immersive experiences, making visible the technical and procedural conditions underlying their construction.
This phase contributes to the development of the Immersion Intensity and Plasticity Scale (IIPS) by moving it from a largely conceptual framework into contact with lived immersive practice. Earlier stages of the research established that immersion and awe can be structured symbolically through narrative, atmosphere, and archetypal patterning. February extends this by testing such conditions within spherical, frameless environments experienced from within rather than imagined in theory.
Dissemination / Practice Ecology
The project operates across literary, visual, and technical domains, positioning itself within a broader ecology of immersive art, cyberdelic experimentation, and practice-based research.
Feedback / Peer Exchange
Video documentation of participants engaging with the immersive environments was captured during informal sessions.
Consent was obtained by all parties for the filming of this informal test group and its inclusion into public blogs. The sessions involved being introduced to VR content for the first time. Three different types of VR media were experienced. ‘Fractal Worlds’ by Julius Horsthuis, ‘Space Explorers’ by Felix & Paul, and then ‘Oblatio’. I made notes of bodily responses and vocalisations in these informal sessions to test the general viability of thematic research and comparison of media through subjects experiencing VR.
Next Steps
Refine the stability and visual coherence of 360° outputs, improve integration between splat environments and video sequences, and expand the functionality of the three.js player. Begin testing the pipeline with live-action compositing and explore potential migration into real-time engines such as Unreal.
Footnotes
- Victor Turner, The Ritual Process (1969).
- Dacher Keltner and Jonathan Haidt, “Approaching Awe” (2003).
- Susan Leigh Star, “The Ethnography of Infrastructure” (1999).
Bibliography
Grau, Oliver. Virtual Art: From Illusion to Immersion. MIT Press, 2003.
Turner, Victor. The Ritual Process. Aldine, 1969.
Keltner, D., and J. Haidt. “Approaching Awe.” Cognition and Emotion 17, no. 2 (2003): 297–314.
Chirico, A., and D. B. Yaden. “Awe.” Current Opinion in Psychology (2022).
Latour, Bruno. Pandora’s Hope. Harvard University Press, 1999.
Star, Susan Leigh. “The Ethnography of Infrastructure.” American Behavioral Scientist (1999).
Tags: #immersivemedia #practicebasedresearch #VR #awe #ritualaesthetics #spatialaudio #worldbuilding #doctoralresearch