Docomo’s AI Research Reassembles Concert Footage Around a Viewer’s Favorite Performer

NTT Docomo says it has developed AI technology that automatically assembles concert footage around an individual viewer’s preferences. Announced on October 23, the research combines close-ups of a chosen performer with shots showing the wider stage and other highlights.

The system uses a vision-language model, which processes visual information alongside language. Unlike a multi-angle stream that asks viewers to switch cameras themselves, the proposed system makes the shot selections automatically. The goal is to retain the broader performance while emphasizing a favorite participant.

Docomo plans a demonstration at the NTT R&D Forum on November 18–20 and November 24–25 at NTT’s Musashino research center. It will combine the video system with NTT audio technology that can emphasize a particular singing voice and control the apparent direction of sound.

This is a research announcement and planned demonstration. The release does not announce a consumer subscription, price or public commercial launch date. Applications to sports and viewing through extended-reality glasses remain areas the company says it will explore.

ドコモ、推しのアップと全体演出を組み合わせるライブ映像技術を発表

ドコモは10月23日、視聴者が指定した出演者のアップとステージ全体の映像をAIで自動構成する研究技術を発表。視覚と言語を処理するVLMを使い、手動のカメラ切替を不要にする仕組み。11月18～20日・24～25日のNTT R&Dフォーラムで音響技術と組み合わせたデモを予定する。一般向けサービスの価格や提供開始日は発表していない。

出典：NTT Docomo official release 773e40ffcce23f31fef35711 via PR TIMES