CAMB.AI and Broadcom embed Voice AI on chip for local translation, dubbing, captioning
When you’re in a Zoom call or watching a live stream, even a half-second pause between your voice and a translation feels oddly disorienting. That’s the kind of lag CAMB.AI is trying to cut out. The Seattle-based startup, which builds speech-tech tools, has just partnered with Broadcom to bake its language-processing stack straight into Broadcom’s newest chip family.
Because the crunching happens on the silicon itself, the audio no longer has to hop to a cloud server and back - a round-trip that can add a few seconds of delay. It also means the raw sound stays on the device, which seems to matter more as privacy rules get stricter. From a cost perspective, streaming services and creators might save a noticeable chunk of their compute budget by dropping the external processing step.
In practice, the deal promises faster, more private, and cheaper multilingual audio. Akshat Prakash, co-founder and CTO of CAMB.AI, summed it up: “By partnering with Broadcom, we can del…”.
The integration enables real-time translation, dubbing, captioning, and audio descriptions to function locally, eliminating latency and privacy concerns while reducing costs for users and content providers. Akshat Prakash, co-founder and CTO of CAMB.AI, said, "By partnering with Broadcom, we can deliver this capability to consumers globally in a way that is faster, more private, and more integrated into everyday devices than ever before." Running on Broadcom's SoC-integrated NPU, CAMB.AI's text-to-speech model converts written text into natural speech in multiple languages. This approach supports accessibility for visually impaired users, improves communication in e-learning and customer service, and cuts reliance on external servers. Rich Nelson, SVP and GM of Broadcom's broadband video group, said, "We are enabling next-generation user experiences that are both highly intelligent and privacy-first." The next phase of the collaboration will explore moving CAMB.AI's real-time translation model to Broadcom's on-device NPU, enabling translation across more than 150 languages.
Local voice AI might start to edge out the cloud, but it’s still a work in progress. The CAMB.AI-Broadcom tie-up slips the MARS generative voice model straight onto Broadcom’s NPU chips, so text-to-speech and localisation can run right on a TV or speaker. Doing translation, dubbing, captioning and audio description locally cuts latency and eases privacy worries.
Akshat Prakash, co-founder and CTO, says the on-chip approach should also lower costs for both users and content providers. The piece, however, leaves out any hard performance numbers, so we can’t tell yet if the on-device quality will really match what the cloud delivers. Because the tech piggybacks on existing hardware, manufacturers could roll it out quickly, provided they’re willing to add the chip.
Cost savings sound plausible, but they’ll probably differ a lot between devices and use cases. All in all, this feels like a tangible step toward offline voice AI, yet its market impact remains fuzzy. We’ll be watching early demos to see how reliable it is and whether people actually like it.
More in Business & Startups
AI shifts hiring: companies favor generalist engineers over specialists
Why are hiring managers rethinking their talent playbooks? The rise of AI tools is nudging firms...
Rivian showcases new AI chip and silicon inspection labs in office tour
Rivian is opening its doors to a behind‑the‑scenes look at the hardware that could define the next...
Anthropic's AI store profits as CEO acknowledges eternal transcendence
Anthropic’s new AI storefront is finally turning a profit, a milestone that feels almost ceremonial...
Gnani.ai launches Vachana STT model as core IndiaAI infrastructure
Gnani.ai, the Bengaluru‑based speech‑technology firm, has just put its newest speech‑to‑text engine...
PC Company Invests in AI Engines, Context‑Aware Computing, Intelligence
A PC maker long known for assembling laptops and desktops is now positioning itself as a backbone...
Common Questions Answered
How does embedding CAMB.AI's voice‑AI on Broadcom's chip reduce latency for real‑time translation?
By integrating the language‑processing stack directly into Broadcom's NPU chipsets, the translation workload is handled locally on the device. This eliminates the need for round‑trip communication with cloud servers, dropping latency to near‑instantaneous levels.
What privacy advantages does the on‑chip MARS generative voice model provide compared to cloud‑based services?
Processing speech data on the device means audio never leaves the user's hardware, preventing exposure to external networks. Consequently, personal conversations and content remain private, reducing the risk of data breaches associated with cloud transmission.
In what ways can the CAMB.AI‑Broadcom integration support dubbing and captioning for live broadcasts?
The integrated voice‑AI can perform text‑to‑speech and localisation directly on the SoC, enabling real‑time dubbing of spoken language and automatic generation of captions. This on‑chip capability ensures synchronized audio and visual outputs without the delays typical of remote processing.
How might the on‑chip solution lower costs for content providers and end users?
Since the heavy computational work is performed locally, there is less reliance on expensive cloud infrastructure and bandwidth. This reduction in server usage and data transfer translates into lower operational expenses for providers and cheaper services for consumers.