CAMB.AI is introducing its new Streaming SDK and Streaming Dashboard at IBC2026, with integrations into NVIDIA Holoscan for Media open reference architecture
The release gives developers and broadcast teams new ways to integrate, manage and deploy real-time multilingual broadcasting within their existing workflows.
Multilingual broadcasting, now live as an SDK
Available for C++, Python, Java and Rust, the SDKs allow developers to start broadcasting in multiple languages with as little as three lines of code, making it easier to integrate multilingual capabilities directly into existing applications and workflows.
CAMB.AI’s live streaming platform supports 150+ languages, producing dubbed audio and subtitles for live content. It supports source-speaker voice cloning, studio voices and custom terminology, allowing names, venues, sponsors and other specific terminology to remain consistent across languages.
The release features the latest quality enhancements backed by CAMB’s foundational models MARS and BOLI.
Bringing live dubbing to NVIDIA Holoscan for Media open reference architecture
CAMB.AI’s new capabilities will be available within the NVIDIA Holoscan for Media reference architecture as a hybrid SaaS.
Holoscan for Media is NVIDIA’s open reference architecture and developer toolkit for building AI-powered live media solutions on NVIDIA infrastructure.
Program audio can be processed by CAMB.AI’s models to produce dubbed commentary in each target language, with outputs returned to the production network or delivered directly to destinations including CDN, OTT and YouTube Live.
For broadcasters, this allows live localization to operate alongside other production applications. It reduces the need to send feeds to a separate cloud localization environment and back, reducing infrastructure hops while allowing media to remain within the broadcaster’s facility or own cloud environment when required.
New Streaming Dashboard
The new Streaming Dashboard provides a visual environment for building, launching and monitoring multilingual streams.
Operators can configure sources, select audio, set processing profiles, create language pipelines and manage output destinations from a single interface. During a live broadcast, teams can monitor stream health, outputs, runtime and live transcripts, as well as pause and resume dubbing without stopping the stream.
CAMB.AI’s live localization technology has already been used across Ligue 1+, NASCAR and Eurovision Sport, including live AI-dubbed commentary and multilingual subtitling.
See It at IBC2026
CAMB.AI will preview its new dashboard and SDKs at IBC2026. Visitors will be able to see live multilingual dubbing in action at the CAMB.AI booth.
CAMB.AI builds localization AI for content, entertainment and sport.
Its MARS family of speech models and livestreaming platform deliver dubbing, translation and subtitles in more than 150 languages, live and on demand.
Founded in 2022, CAMB.AI has offices in Dubai and San Francisco.
Facts Only
* CAMB.AI is introducing a new Streaming SDK and Streaming Dashboard at IBC2026.
* The SDKs support C++, Python, Java, and Rust for multilingual broadcasting implementation.
* The live streaming platform supports over 150 languages, producing dubbed audio and subtitles.
* The system supports source-speaker voice cloning, studio voices, and custom terminology consistency across languages.
* Quality enhancements are backed by CAMB’s foundational models MARS and BOLI.
* Capabilities will be integrated into the NVIDIA Holoscan for Media open reference architecture as a hybrid SaaS.
* Program audio can be processed to produce dubbed commentary in target languages delivered to production networks or destinations like CDN, OTT, and YouTube Live.
* The technology allows live localization to operate alongside other production applications, reducing infrastructure hops.
* A Streaming Dashboard allows operators to configure sources, set profiles, create pipelines, and monitor streams live.
* CAMB.AI’s technology has been used for live AI-dubbed commentary in Ligue 1+, NASCAR, and Eurovision Sport.
* CAMB.AI’s MARS family of speech models and livestreaming platform handle dubbing, translation, and subtitling.
* CAMB.AI was founded in 2022 and has offices in Dubai and San Francisco.
Executive Summary
Full Take
The convergence of sophisticated AI models (MARS, BOLI) with established media infrastructure (NVIDIA Holoscan for Media) presents a specific tension between novel capability and real-world deployment logistics. The value proposition hinges on abstracting complex localization tasks—dubbing, translation, terminology management—into accessible SDKs for existing broadcast workflows. This move attempts to democratize high-level media processing, moving it from bespoke cloud environments into integrated reference architectures.
The pattern emerging is the strategic layering of AI services onto established hardware and industry standards. The integration with Holoscan suggests an attempt to standardize a previously fragmented localization pipeline, aiming to reduce latency and infrastructure complexity for broadcasters who prefer maintaining control within their own facilities or private clouds. This addresses a known friction point in live localization: the need for low-latency processing without constant back-and-forth data transfers.
The existence of historical use cases across major sports events suggests that the underlying technology has already passed certain robustness tests in high-stakes environments, but the narrative surrounding the future must guard against overstating the seamlessness of this integration. The risk lies in treating infrastructure as purely a delivery mechanism rather than a system requiring deep operational understanding. The question for observers is whether the focus on ease-of-use via SDKs obscures the underlying complexity of maintaining multilingual consistency and quality across diverse, real-time production environments. What mechanisms are in place to ensure that custom terminology or nuanced context learned through voice cloning remains faithfully preserved when shifting between different hardware/cloud delivery modalities?