Artificial intelligence news.
Straight from the original announcements.

VoxNexus launches offline Android voice AI that listens 12 hours on 2.49% battery

October 23, 2026, 18:31Source: PR Newswire

VoxNexus has released VoxEdge, an on-device voice stack that brings speech recognition, translation and speech synthesis to affordable Android devices without any internet connection. The company also launched VoxSense, an ambient listening solution that, according to VoxNexus, uses 162 mAh for 12 hours of continuous listening and sends no audio to the cloud.

Key points
  • VoxNexus released VoxEdge, an offline on-device voice stack for affordable Android devices.
  • VoxEdge supports speech recognition in 36 languages, translation in 103 and speech synthesis in 45.
  • VoxSense uses 162 mAh for 12 hours of continuous listening, around 2.5% of a 6,500 mAh battery, according to the company.
  • Both products were unveiled at the MediaTek Dimensity 9600 launch in Madrid on October 19.
  • Developers can integrate via a license and SDK and request a free trial.

VoxEdge: speech, translation and synthesis offline

According to VoxNexus, VoxEdge supports speech recognition in 36 languages, translation in 103 languages, and speech synthesis in 45 languages. Everything runs on the device, with no connection, no data bundle and no cloud token bills.

The company says every query is free once the model is installed. Its speech recognition model has as few as 65 million parameters and, according to VoxNexus, achieves a lower word error rate than open-source models nearly ten times its size.

The release was unveiled at the MediaTek Dimensity 9600 launch in Madrid on October 19. There, VoxNexus demonstrated live on-device speech recognition: Spanish and English conversations were transcribed in real time on the phone, with no cloud connection.

The problem VoxNexus says it is addressing

VoxNexus argues that AI should work for everyone, not only for speakers of a handful of mainstream languages. The company says voice AI across much of Africa has been held back by scarce training data and a lack of targeted fine-tuning. Everyday speech also includes code-switching and dialect variation, and cloud-based inference can make deployment at scale expensive.

"Under-resource languages aren't small languages. Hundreds of millions of people speak them. Until now, AI models just haven't learned them yet," said Charles Shi, CEO of VoxNexus.

VoxSense: always-on listening with a small battery cost

VoxSense is an on-device ambient listening solution. VoxNexus says it uses only 162 mAh for 12 hours of continuous listening, around 2.5% of a 6,500 mAh phone battery. The headline figure in the company's announcement is 2.49%.

Because all processing stays on the device, the company says always-on listening becomes practical without draining the battery and without sending any audio to the cloud.

Developers and device makers can integrate it through a license and a simple SDK. Developers can request a free trial via the VoxNexus website, www.VoxNexus.ai.

What it means for developers and device makers

For developers and device makers, the pitch is voice features that work offline, keep audio on the phone and carry no per-query cloud cost. VoxNexus describes its audio intelligence as delivering low latency, built-in privacy and zero token cost on any device with a chip. The company is headquartered in Singapore.

The performance and battery figures come from VoxNexus itself, and the announcement does not include independent test results or pricing for the license.

VoxNexusVoxEdgeVoxSenseMediaTek Dimensity 9600voice and speechsmartphonesprivacy

Related news

Meta Brings Real-Time Voice Conversations to Muse
Products and toolsNovember 5, 2026, 19:30

Meta Brings Real-Time Voice Conversations to Muse

Meta is rolling out real-time voice conversations with Muse, its AI assistant, across the Muse app on mobile, web and Mac. Users can speak with their Muse, interrupt…

Hugging Face introduces AnySpeechRecognizer and AnyDecisionModel for Apple platforms
ModelsOctober 29, 2026

Hugging Face introduces AnySpeechRecognizer and AnyDecisionModel for Apple platforms

Hugging Face has introduced AnySpeechRecognizer and AnyDecisionModel, which its blog describes as open models for speech and decisions on Apple platforms. The post…

Siri AI expands to French, Japanese, Korean, Portuguese and Spanish
Products and toolsOctober 26, 2026

Siri AI expands to French, Japanese, Korean, Portuguese and Spanish

Apple has expanded Siri AI to five more languages with its 27.2 software updates: French, Japanese, Korean, Portuguese and Spanish. According to Apple, more languages…