Document Processing
Pacers Sports & Entertainment needed a custom captioning system that could keep up with fast, unscripted basketball commentary and deliver accurate live captions and translations to deaf, hard-of-hearing, and non-English-speaking fans on arena displays and in its mobile apps.
Reported impact
PS&E launched the first live in-arena and mobile captioning experience of its kind for Pacers and Fever games, reduced caption error to 1.14%, cut model-training preparation time from days to minutes, and created a scalable foundation for multilingual and future AI-powered fan experiences.
Pacers Sports & Entertainment built a real-time captioning stack on Microsoft Azure to turn live game commentary into synchronized captions across Gainbridge Fieldhouse screens and the Pacers and Fever mobile apps. Azure AI Foundry provided a secure, containerized environment for training and refining custom Azure AI Speech models with hundreds of hours of Pacers and Fever broadcasts plus structured lists of player, coach, and official names. In production, announcer audio is transcribed and translated by Azure AI Speech, then Azure Functions processes caption events and Azure SignalR Service pushes them to a single endpoint used by both in-arena displays and the mobile app with virtually no lag. The architecture gave PS&E near-instant English captions, added Spanish, and created a path to roll out 12 additional languages and more AI-powered fan experiences.
Live announcer audio is sent into Azure AI Speech models trained in Azure AI Foundry, where it is transcribed and translated in real time. Azure Functions processes the caption output and Azure SignalR Service publishes it to a shared endpoint that keeps the Pacers and Fever mobile apps and Gainbridge Fieldhouse displays in sync.
"By tailoring the model to the Pacers’ broadcast style, they reduced the speech-to-text transcription error rate by 87%."
"Captions are delivered through an event-driven architecture using Azure Functions and Azure SignalR Service."
"Originally, the audio file for a single game took us upwards of five days to manually cut the game audio. Now, using our custom script, it takes us roughly two minutes to cut and encode a game and unload for translation."
"The system had to follow fast, unscripted announcers, recognize game-specific lingo and player names, and still deliver captions instantly to keep fans engaged in real time."
"With each iteration, the team pushed well past their original 10% error rate goal, landing at just 1.14%."
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Microsoft Sample Apr 26, 2024 MultiChoice Like many, MultiChoice’s IT Project Management Office (PMO) struggled with managing a day-long chain of back-to-back meetings and minutes. Embracing Microsoft Copilot, MultiChoice’s IT PMO automated meeting minutes,… "Previously, we aimed for minutes to be sent out within 24 hours of a meeting. Microsoft Copilot · Microsoft Teams Microsoft Sample Oct 15, 2026 Bar-Ilan University Students who miss lectures struggle to reconnect new lessons with prior material when course information is spread across notes, recordings, slides, and readings, and lecturers need a reliable way to answer routine… 71 courses Azure AI Services · Microsoft Azure · Microsoft Copilot Studio 
OpenAI Sample Oct 8, 2026 Axios AI is already a huge part of how Axios Local works. At the core, what we’re trying to do is prove that you can run a sustainable, profitable local news model that delivers high-quality journalism to every community in… Axiomizer · custom GPTs 