Preserving cultural heritage: Inside Google DeepMind’s collaboration with Pelé

Source: Google Date: 2026-06-23 URL: https://blog.google/company-news/inside-google/around-the-globe/google-europe/preserving-cultural-heritage-inside-google-deepminds-collaboration-with-pele/

Summary

Google DeepMind partnered with historians, football journalists, surviving contemporaries, and Pelé’s family to reconstruct the “Gol da Rua Javari” — a 1959 goal widely regarded as the most beautiful of Pelé’s career that was never captured on film. The reconstruction used Gemini Omni to process historical records and period accounts, and Veo 3 to generate the visual footage, with authentic details sourced from the original pitch, period uniforms, and a vintage ball. The resulting reconstruction will be exhibited at the Pelé Museum.

Implications

• Model capability: This is a practical demonstration of Veo 3 at the quality threshold required for institutional display — the Pelé Museum setting functions as a curated review of the output’s fidelity. More useful as a capability signal than a benchmark score because the stakes (family approval, historical accuracy, museum installation) are real.
• Voices/power dynamics: The choice to reconstruct a Black Brazilian football legend’s lost moment using a US tech company’s video generation model raises implicit questions about who controls the aesthetic and historical record. The collaboration with family and historians is the answer Google provides; whether that’s sufficient governance is a legitimate debate.
• Agent landscape: Gemini Omni handling historical source synthesis while Veo 3 handles generation is an early template for multi-modal, multi-model pipelines applied to creative and cultural work — distinct from the developer-tooling agent patterns but worth tracking as the same underlying architecture.