MIT alumni Serena Pei and Josephine Lee, with backgrounds in MIT CS and experience at Scale AI, Datadog, and Amazon, launched Palette as a two-person team in the YC S26 batch, with Pei as CTO and Lee as CEO. Palette focuses on a unified generation engine and model routing covering Seedance, Veo, Kling, and others, rather than a single model. The product aims to address brand consistency pain points for enterprises and the fragmentation of creator tools, upgrading one-time generation to a repeatable workflow through natural language and contextual memory, targeting high-frequency scenarios in L&D, advertising, and product content. Compared to early video generation tools like Runway, Pika, and Luma, which remain at "one-time generation," Palette is positioned as a "full-process operating system," marking a shift from model capability demonstration to enterprise-level workflow reconstruction. The approach replaces the traditional "prompt → export → multi-software stitching" chain with a single dialogue engine, shifting pricing power from the number of tools and labor hours to contextual consistency and routing efficiency.