CAMB.AI, a global multilingual voice translation company, has partnered with Kompact AI, Bengaluru-based deep tech Ziroh Labs’ CPU-first AI platform. With the partnership. CAMB.AI aims to run its multilingual voice and translation models on regular CPU machines and remove the need for expensive GPU systems.
The UAE- and US-based company will utilise Kompact AI’s optimisation technology to run its CAMB.AI’s MARS7 (Multilingual Audio Rendering and Synthesis) and BOLI (Bilingual Online Language Interpreter) models. Kompact AI also supports real-time use on edge devices, including offline environments, enabling organisations to scale their voice AI systems using existing hardware.
CAMB.AI’s co-founder and CTO, Akshat Prakash, said the collaboration ensures enterprises “no longer need to choose between performance and accessibility,” adding that companies can now deploy MARS7 and BOLI models on standard CPU infrastructure while achieving broadcast-quality results across 150+ languages.
With the partnership, CAMB.AI looks to reduce AI inferencing costs by up to 50% while eliminating the need for quantisation or distillation, which can reduce accuracy.
Hrishikesh Dewan, co-founder and CEO of Ziroh Labs, highlighted that the collaboration creates “unprecedented opportunities for enterprises to deploy sophisticated AI without traditional barriers” and described the CPU-first runtime as a potential game-changer across industries from sports to education to healthcare. Kompact AI enables large language models to perform high core counts on CPU infrastructure, challenging the traditional GPU-centric model.
The integration supports a wide range of real-world applications, the company noted in a statement. Sports venues can run live multilingual commentary locally without relying on the cloud.
Hospitals can operate translation systems on-premises while ensuring regulatory compliance. Educational institutions can provide multilingual learning using their existing infrastructure. Global enterprises can enable real-time translation in meetings and customer service centres without expensive hardware upgrades.
With CPU-optimised MARS7 and BOLI models, CAMB.AI now looks to make scalable, high-quality multilingual AI accessible across industries, removing barriers that previously limited enterprise adoption.
Last month, CAMB.AI and Broadcom announced a collaboration that embeds CAMB.AI’s generative voice model, MARS, directly into Broadcom’s neural processing unit chipsets.
CAMB.AI has worked with brands including NASCAR, Major League Soccer, YES Network, the Australian Open, FanCode, and Comcast NBCUniversal to enable emotionally authentic multilingual dubbing for live sports.
The company has raised $18.3 million to date, according to Tracxn data, and operates across the US, Canada, APAC, Europe, the Middle East, and India markets. Its MARS generative text-to-speech model is available on AWS Bedrock and Google Vertex AI.
The post CAMB.AI, Kompact AI by Ziroh Labs to Optimise Voice Synthesis LLMs for CPUs appeared first on Analytics India Magazine.
