Show HN: Spica – OSS Tool to Generate Infinite Length Sora-2 Videos (spica.kuber.studio)

Spica is an open-source utility by Kuber Mehta that orchestrates the Sora-2 video model to produce effectively infinite-length videos by generating and stitching segments with continuity. Users configure a generation pipeline, upload a starting image as the seed for the first segment, and Spica iteratively produces subsequent segments so the output can extend beyond typical model-length limits (the demo shows a ~16s total duration stub). The interface focuses on preserving temporal coherence across boundaries rather than producing standalone clips. This matters for the AI/ML community because long-form video generation raises different technical challenges than single-shot outputs: maintaining consistent scene state, preventing visual drift, and managing compute/memory over long horizons. Spica provides a practical, open framework to experiment with techniques for continuity (e.g., seeding future segments from previous frames or latent states), pipeline configuration, and stitching strategies. That makes it useful for researchers and creators exploring persistent characters, background loops, or interactive experiences, while also surfacing important failure modes (artifact accumulation, narrative inconsistency, and resource scaling) that merit further model- and systems-level work.