Gather25, hosted by IF:Gathering, was a groundbreaking 25-hour global broadcast connecting audiences across every continent. With over 1.25 million online viewers and participation from more than 21,000 Gather Groups worldwide, the event aimed to inspire and unite the global church community.
Aberdeen Broadcast Services has partnered with IF:Gathering since 2020, providing both live captioning for streaming events and post-produced captioning for archived content. Gather25 marked an ambitious new chapter in that partnership—the first time Aberdeen was brought on to help scale the event’s global reach through real-time translation.
As far as we know, this was a first-of-its-kind undertaking: a multilingual livestream of this magnitude, requiring impeccable coordination across dozens of languages and platforms. Aberdeen was one of several trusted vendors, working alongside the technology and broadcast partners hired for the event to make it all possible.
The Challenge: Delivering 25 Hours of Multilingual, Real-Time Access
Supporting a continuous, multilingual broadcast of this scale introduced several technical challenges:
Maintain uninterrupted captioning across a 25-hour live stream spanning 7 global broadcast hubs.
Generate real-time captions and translations for 84 language streams with high accuracy and minimal latency.
Implement a dynamic captioning system that allows viewers to pause, rewind, and review up to four hours of content during the live event ("rolling DVR" functionality).
Collaborate across multiple teams — Element Media Group (master control), Sardius (video player delivery), SyncWords (captioning/subtitle platform), and Interprefy (live interpreters) — to ensure flawless live execution.
Deliver caption files in a continuous VTT format, dynamically updating during the broadcast to support both live viewing and post-event on-demand access.
The Solution: Continuous Live Captioning and Translation Across 84 Languages
Before the event went live, a workflow that could support uninterrupted, real-time captioning and translation across dozens of simultaneous streams needed to be built. This required deep coordination with multiple teams to align audio sources, language feeds, and delivery endpoints. Our goal was to ensure every segment of the broadcast could be accurately captioned and translated with minimal manual intervention once the event began.
A key technology partner behind this workflow was SyncWords, whose platform Aberdeen leverages to manage real-time captioning and translation delivery. SyncWords played a vital behind-the-scenes role in not only powering the infrastructure we used to deploy captions across dozens of streams but also collaborating directly with engineers at Sardius and Elemental Media to implement specialized audio-isolation coding for the event. This coordination ensured that every audio feed we received was optimized for clean, accurate transcription and translation at scale.
Pre-Event Testing
To guarantee performance, Aberdeen conducted over 50 hours of pre-event testing, stress-testing multi-language streams, and simulating 25-hour sessions to ensure system endurance. A key priority was ensuring that VTT caption files would function as continuously updated feeds, rather than static uploads. This real-time updating was essential for Sardius’ platform to support both live captions and rolling DVR features, while also allowing seamless access to captions during on-demand playback after the event.
Streaming Caption Workflow
In a typical video workflow, captions are created after recording, uploaded separately, and synced to on-demand content. For Gather25’s livestream, captions had to be generated and delivered in real time, alongside the video stream.
The system worked like this:
Live video was broken into segments (HLS .ts files) and stored in an S3 bucket, alongside a .m3u8 manifest file managing playback.
Simultaneously, a VTT subtitle file was created and appended in real time—each new caption line was added to the file without restarting it.
This continuous file structure allowed Sardius’ player to provide viewers with immediate access to live captions while also enabling DVR controls—viewers could rewind up to four hours and then return to the live stream without losing their place.
Once the event ended, these same VTT files remained attached to the recorded content for full accessibility during on-demand viewing.
This architecture required precise timing, structured file delivery, and full alignment with Sardius’ streaming infrastructure.
What's an HLS Stream?
An HLS stream (HTTP Live Streaming) delivers video content over the internet in small, manageable chunks. The video is split into short segments (usually 2–10 seconds long) and saved as .ts (transport stream) files. A playlist file (called a .m3u8) tells the video player what order to play those chunks in. As a viewer watches, their device downloads and plays the segments one at a time, allowing smooth playback, even with slow or fluctuating internet.
Workflow Optimization
Aberdeen’s team engineered a sophisticated workflow to process 20 incoming HLS feeds. Working closely with Element Media Group, which managed master control and delivered stripped audio to Sardius, Aberdeen received feeds prepped for accessibility:
Speech-Only Audio Extraction: Each of the 20 HLS streams included up to 16 audio tracks; Aberdeen pulled content from Track 16, where music had been removed to isolate spoken word, ensuring clean inputs for accurate captioning and translation.
Language Processing:
Live English captioning was performed by Aberdeen’s professional human captioners via a broadcast encoder at Master Control, ensuring exceptional accuracy of the English captions for the television broadcast.
Translated English to Kinyarwanda, a language not supported by any of the major ASR engines, demonstrating a custom solution beyond standard automation.
Created native-language captions for 14 non-English spoken feeds, including Arabic, Mandarin, French, German, Portuguese, and more.
Translated stripped English audio into 68 other languages, selecting the best available machine translation engine per language (prioritizing: DeepL > OpenAI > Amazon > Google > Azure).
File Management: All caption files were delivered to Sardius using clearly labeled S3 bucket destinations (e.g., EN for English, ES for Spanish), with structured naming conventions to ensure accurate mapping during the live event and in the post-event archive.
Impact: Real-Time Engagement and Global Reach Through Accessibility
Gather25’s accessibility efforts produced measurable results:
Over 1 million minutes of English captions and translations were viewed during the live stream, equal to nearly 17,000 hours of real-time engagement.
Continuous, real-time VTT streaming ensured seamless live accessibility, enabled four-hour rolling DVR, and extended caption access post-event.
Aberdeen's workflow operated without any critical interruptions across the entire 25-hour window, validating the strength of its testing protocols and system reliability.
By combining AI-driven automation, human captioning expertise, and a deep integration with broadcast systems, Aberdeen Broadcast Services delivered scalable, high-quality accessibility at a truly global level.
Conclusion: Advancing Live Event Accessibility on a Global Scale
Gather25’s mission to unite believers around the world was made stronger through its commitment to accessibility. With real-time captioning and translation across 84 language streams, Aberdeen Broadcast Services helped make this global event inclusive, impactful, and available to all.
For organizations planning large-scale, multilingual broadcasts, Aberdeen’s tested and proven solutions offer the reliability and scalability needed to reach a worldwide audience.
Let’s talk about how we can support your next event.Contact us to learn more.
