Google Antigravity has developed an AI translator that functions without the need for an internet connection.

      **Google**

      Google has unveiled a new AI initiative that could make language barriers less daunting, and it operates without needing an internet connection. The company has launched the Gemma Translator, a compact prototype developed in partnership with Antigravity. Unlike most AI translation tools that depend on cloud processing, this device functions entirely offline with the Gemma 4 E2B, Google's lightweight open model. All processing occurs locally on the device, making it portable and independent of internet access. The prototype is powered by a Raspberry Pi 5 and features a speaker and microphone housed within a custom 3D-printed enclosure, resulting in a self-contained translator that is easy to carry.

      **AI translation without relying on the cloud**

      The demonstration highlights the advancements made in AI models. Instead of depending on external servers, the Gemma Translator processes spoken language directly on the Raspberry Pi, translating conversations in real time.

      Meet the Gemma Translator! A fully offline device powered by Gemma 4 E2B, developed in collaboration with @Antigravity. Running solely on a Raspberry Pi 5 with an integrated microphone and speaker, this highly mobile prototype is encased in a custom 3D-printed shell. pic.twitter.com/GhqlLC1RuE— Google Gemma (@googlegemma) August 6, 2026

      This approach offers several clear benefits. Travelers could utilize it in areas with spotty connectivity, while businesses or organizations engaged in sensitive discussions wouldn't need to upload voice data to the cloud. Running everything locally could also enhance responsiveness, making conversations feel more fluid. While Google has not positioned this project as a commercial offering, it showcases the possibilities now available with effective AI models that can function on affordable hardware.

      **A preview of what's to come**

      The Gemma Translator is more of a technological showcase than an actual product reveal. However, it indicates a future where AI assistants may not always require powerful data centers to be effective.

      As open models keep shrinking while gaining capability, devices like this might eventually be integrated into travel gadgets, accessibility solutions, educational tools, or even emergency response kits where internet connectivity can't be guaranteed. For now, Google's latest prototype serves as a reminder that some of AI's most fascinating innovations are being compacted into small devices that can operate virtually anywhere.

      Shimul is a contributor at Digital Trends, bringing more than five years of experience in the technology sector.

      **Anker highlights a major issue with portable power stations**

      When purchasing a portable power station, one figure is hard to overlook: battery capacity. A 2 kWh battery should provide approximately 2 kWh of usable power, right? Unfortunately, it's not that straightforward. Anker SOLIX has released new efficiency data for its S Series portable power stations, shedding light on an often-overlooked aspect: how much of the energy stored in the battery actually powers your devices.

      Anker states that portable power stations are typically advertised with efficiency ratings determined under relatively heavy loads, achieving around 89% to 92% efficiency. However, this doesn't reflect how most individuals use them during outages. Consider what you might plug in. A Wi-Fi router might draw power continuously, a refrigerator turns its compressor on and off throughout the day, and a CPAP machine might operate overnight. Collectively, these devices may only use a small fraction of what a large power station can supply, leading to decreased efficiency.

      **OpenAI’s first device resembles a compact, expressive AI companion**

      Reports for months have suggested that OpenAI's inaugural hardware product would be a screen-free smart speaker designed to function more like an AI companion. We already knew it could perceive its environment, be transported around the home, and assist users proactively. A recent Bloomberg report now provides more clarity about the potential design of the device.

      According to the report, OpenAI’s first gadget will have a doughnut-like shape and be roughly the size of a hockey puck. You will be able to carry it between rooms or place it conveniently on any surface. The device is projected to be on the pricier side, with the company considering a price range of approximately $300 to $400. A launch is currently slated for 2027.

      **After using ChatGPT Voice, typing now feels agonizingly slow**

      You already know the routine: it’s a new day, which means I've discovered yet another way to incorporate ChatGPT into my daily life. This time, it's entirely about Voice. I've been using ChatGPT's voice mode almost everywhere. It began with my phone, then transitioned to my laptop and desktop, and yes, I've even found myself chatting with it through Apple CarPlay while driving. What started as an occasional interaction has now become a regular part of my routine.

      The timing couldn't have been better. OpenAI recently launched GPT-Live, its next-generation voice experience, engineered to sound more natural and facilitate much more intelligent conversations. After typing countless prompts, communicating with ChatGPT through voice has been surprisingly refreshing.