Google utilized Veo and Gemini to recreate the most remarkable goal Pelé ever scored, a moment that was never captured on film.

      In short, Google DeepMind has recreated Pelé’s 1959 “lost goal” using Veo, Gemini Omni, and Nano Banana Pro. The project was informed by 2,000 records and interviews with eyewitnesses and is currently showcased at the Pelé Museum.

      On August 2, 1959, Pelé achieved what he considered the best goal of his career: three consecutive sombreros over defenders, a flick past the goalkeeper, and a header into the net, all while the ball never touched the ground. This goal was never captured on film, making the “Gol da Rua Javari” live only in the memories of attending fans until now. Google DeepMind has reconstructed it.

      This initiative, in collaboration with Pelé’s family and the Pelé Brand, integrated traditional filmmaking with three Google AI models: Veo 3, Gemini Omni, and Nano Banana Pro. Historian Anita Lucchesi collected nearly 2,000 historical records, from stadium designs to family photos, and conducted interviews with surviving witnesses. A crew filmed live-action scenes at the original Rua Javari stadium using period-appropriate leather balls and uniforms. The footage was then processed through the AI models, which replaced the stunt performer with Pelé’s likeness, modified the modern stadium to reflect 1959’s architecture and conditions, and created period-appropriate crowd scenes.

      Flávia Kurtz, Pelé’s daughter, remarked, “He would be so proud to see all this happening. He’d always lament that the goal was never recorded.” The reconstruction is now on display at the Pelé Museum in Santos. Gemini Omni was introduced as a conversational video-generation model at I/O 2026, with the Pelé project being its most culturally significant application so far, showcasing that AI video generation can aid in preservation rather than merely content production.

      The final output underwent processing through a filmout machine to replicate the appearance of 1950s cinema, followed by traditional VFX enhancements for ball compositing and color grading. Designed for precise image generation, Nano Banana Pro demonstrated its capabilities in the Pelé reconstruction by not just generating content from scratch but reassembling a real event from various evidence fragments. In a year where AI has often been employed to inundate the internet with synthetic material, this instance stands out as a unique case of utilizing it to recover something genuinely lost.