VoxNexus VoxEdge and VoxSense: On-Device Voice AI in 100+ Languages That Never Sends Audio to the Cloud

 

Voice assistant on smartphone screen

A Singapore-based company called VoxNexus has launched two on-device voice AI products that work entirely offline: VoxEdge, a voice stack for affordable Android devices, and VoxSense, an always-on ambient listening solution that uses only a tiny fraction of a phone's battery. According to a PR Newswire announcement on October 23, 2026, both products were unveiled at the MediaTek Dimensity 9600 launch in Madrid on October 19, and neither sends audio to the cloud. For users in areas with expensive data, weak connectivity, or under-served languages, that combination could matter a lot.

The problem VoxNexus is trying to solve

Most voice assistants today depend on the cloud. You speak, your audio travels to a distant server, the server processes it, and an answer comes back. That model works well in cities with fast, cheap internet, but it breaks down elsewhere. As VoxNexus puts it in its announcement, across much of Africa voice AI has been held back by scarce training data and the lack of targeted fine-tuning, while everyday speech also includes code-switching — mixing languages in a single sentence — and dialect variation. Cloud-based inference costs can also make deployment at scale expensive for device makers.
VoxNexus's answer is to move the whole voice stack onto the device itself. Once the model is installed, every query is free, according to the company, because there is no cloud token bill, no data bundle needed, and no connection required. That is a bold claim, and it is important to note that it is the company's own description of its products, not an independent test result. Still, the technical details give a concrete picture of what is being offered.

VoxEdge: speech, translation and synthesis, all offline

VoxEdge is an on-device voice stack that brings speech recognition, translation and speech synthesis to affordable Android devices, entirely offline, according to the PR Newswire announcement. The headline numbers are: Speech recognition in 36 languages. Translation in 103 languages. Speech synthesis in 45 languages. All running with no connection, no data bundle, and no cloud token bills.
VoxNexus says its speech recognition model has as few as 65 million parameters and achieves a lower word error rate than open-source models nearly ten times its size. A smaller model matters because it can run on cheaper phones with less memory and less power, which is exactly the market VoxEdge targets. The company also highlights support for under-resource languages. As VoxNexus CEO Charles Shi is quoted as saying, "Under-resource languages are not small languages. Hundreds of millions of people speak them. Until now, AI models just have not learned them yet."
For everyday users, the practical promise is simple: a phone that can transcribe speech, translate a conversation, and speak back — even with no SIM card, no Wi-Fi, and no data plan. For device makers, the promise is a voice feature they can ship on affordable hardware without paying per-query cloud costs.
Person using voice assistant on phone


VoxSense: always-on listening on 2.5% of a battery

VoxSense is the more striking of the two products, because always-on listening is normally a battery killer. According to the announcement, VoxSense uses only 162 mAh for 12 hours of continuous listening, around 2.5% of a 6,500 mAh phone battery. The company's headline sums it up: "Hears everything, uploads nothing."
By keeping all processing on device, VoxSense makes always-on listening practical without draining the battery and without sending any audio to the cloud, according to VoxNexus. That distinction matters for privacy: with a cloud assistant, your voice data leaves your phone by design; with an on-device system, it does not need to. Of course, privacy also depends on how a device maker implements the product, what permissions it requests, and whether any data is stored locally — details that will need checking in real devices, not just in a launch announcement.
VoxNexus demonstrated live on-device speech recognition at the MediaTek Dimensity 9600 launch in Madrid on October 19, 2026, according to the announcement, with Spanish and English conversations transcribed in real time on the phone, with no cloud connection. A live demo is stronger evidence than a slide deck, but it is still a controlled demonstration, not a review of a shipping consumer phone.
Smartphone with voice AI concept


How developers and device makers can use it

VoxNexus says developers and device makers can easily integrate its solutions through a licence and a simple software development kit, or SDK, and developers can request a free trial through the company's website. That is an important detail for readers: VoxEdge and VoxSense are not, at this stage, consumer apps you can download today from a store. They are building blocks for phone makers and app developers to integrate into their own products.
That means the timeline for ordinary users is indirect. First, a device maker or developer adopts the stack; then it appears in a phone or app; only then can users try it. The announcement does not name any shipping phone that includes VoxEdge or VoxSense yet, so readers should not expect to find these features on their current phone tomorrow.

Why this launch matters — and what to watch

VoxNexus's launch matters for three reasons. First, it shows that capable voice AI no longer necessarily requires a flagship phone or a cloud connection — a 65-million-parameter model can, on the company's claims, beat much larger open-source models on accuracy. Second, it puts under-served languages at the centre, rather than treating English and a handful of major languages as the only ones that matter. Third, it offers a privacy model — on-device only — that is fundamentally different from cloud assistants.
At the same time, readers should keep three cautions in mind. The accuracy, battery and language claims come from the company itself and have not yet been independently verified in shipping devices. The products are developer tools, not consumer apps, so availability depends on adoption by phone makers. And on-device processing, while more private by design, still needs transparent implementation to earn trust in always-on listening.
If VoxEdge and VoxSense do reach affordable Android phones in the coming year, they could change what voice AI looks like for hundreds of millions of users: no data cost, no cloud delay, and support for languages that today's assistants handle poorly. That is a goal worth watching, even while waiting for independent proof.
Voice assistant speaker and phone illustration


Disclaimer: Product details and performance claims in this post are based on VoxNexus's own announcement via PR Newswire on October 23, 2026; independent reviews of shipping devices are not yet available, and features and availability may change.

Post a Comment

Previous Post Next Post