SSPAI Morning Brief: Sony and Insta360 Push Camera Innovation with 65mm Cinema Imaging and AI-Powered Head Tracking

Morning Brief

Three Google Updates
Huawei Unveils the Vision Smart Screen S7 X Pro
Apple and Google Jointly Launch the Open-Source Eclipsa Video HDR Standard
Sony Announces Development of the RIALTO 65 Image Sensor Module
Insta360 Officially Unveils the Luna Ultra Gimbal Camera Head-Tracking Module
Microsoft Plans to Use AI to Accelerate Windows 11 ARM Ecosystem Compatibility
Apple Store Developer Billings and Sales Surpassed $1.4 Trillion in 2025
News Worth a Quick Look

Three Google Updates

Google Launches the AI App Dreambeans

On June 4, Google Labs introduced a new AI-powered app called Dreambeans. With user permission, the app integrates information from personal Google services and generates a limited number of daily lifestyle inspirations, encouraging users to put down their phones and engage more with real life. The app is currently available only to eligible Google AI Ultra subscribers in the United States on Android and iOS, while users with personal Google accounts can join a waitlist.

According to Google, Dreambeans is built on personal intelligence technology and can integrate data from Gmail, Google Calendar, Google Photos, YouTube, Search history, and other Google services after receiving user authorization. Based on a user’s interests and preferences, it generates between 10 and 14 curated lifestyle notes or information snippets each day.

Dreambeans can recommend content based on a user’s activities and interests. For example, it may suggest a newly opened café near the user’s home that matches their preferences. If a user has an upcoming puppy adoption recorded in Google Calendar, Dreambeans may surface educational content about raising young dogs. The app can also curate articles from across the web that align with the user’s interests. Source

Google Releases and Open-Sources the Gemma 4 12B Multimodal Model

On June 4, Google announced and open-sourced the Gemma 4 12B multimodal model, aiming to make local AI inference accessible on consumer-grade devices. According to Google, internal testing shows that the model can run on laptops and desktop computers equipped with as little as 16GB of system memory or VRAM.

Gemma 4 12B features 12 billion parameters and maintains a relatively compact model size while delivering intelligence levels comparable to the larger Gemma 26B model. This design makes it particularly suitable for deployment on local devices, development environments, and edge-computing scenarios, lowering the hardware requirements for developers working with multimodal AI.

On the multimodal front, Gemma 4 12B simplifies the handling of image, video, and audio inputs. Unlike traditional multimodal models that often rely on separate encoders to convert images or audio before passing them to the language model, Gemma 4 12B integrates input processing more directly, helping reduce deployment and operational complexity.

The model is currently available across multiple platforms. Developers can try it directly through tools such as Ollama or download model weights from Hugging Face and Kaggle. Google also noted that developers can use Unsloth for efficient fine-tuning, enabling customized versions tailored to specific use cases. Source

Google AI Edge Gallery Arrives on macOS

On June 4, Google officially released a macOS version of its on-device generative AI application, Google AI Edge Gallery. According to Google, the app focuses on offline operation, privacy protection, and low-latency performance. Since models run locally on the user’s device, no processing needs to be uploaded to the cloud, making it suitable for scenarios where data privacy and responsiveness are especially important.

Google currently provides access to five models within AI Edge Gallery: Gemma-4-12B-it, Gemma-4-E2B-it, Gemma-4-E4B-it, Gemma-3n-E2B-it, and Gemma-3n-E4B-it. Users can choose different models based on device performance and specific needs to experience generative AI capabilities locally.

At the same time, Google AI Edge Eloquent has also arrived on macOS. The application is a free dictation tool that captures spoken input and converts it into text while removing verbal fillers and lightly refining the output to improve clarity. Google emphasized that Eloquent’s processing is performed entirely on-device rather than in the cloud. Source

Huawei Unveils the Vision Smart Screen S7 X Pro

Huawei has listed the Vision Smart Screen S7 X Pro on its official website. The TV is available in 65-inch, 75-inch, and 85-inch sizes, with pricing starting at RMB 6,299. Huawei has not yet announced an official release date.

In terms of specifications, the Vision Smart Screen S7 X Pro features a Super MiniLED display with over one thousand local dimming zones and a resolution of 3840 × 2160. In Game Mode, the panel supports an effective refresh rate of up to 288Hz through frame interpolation technology and includes Huawei’s AI-powered precision backlight control algorithm. For audio, the TV is equipped with a three-way 2.1-channel surround sound system, with bass extending down to 60Hz. The device ships with HarmonyOS 4.3 and supports memory expansion technology.

Feature-wise, the Vision Smart Screen S7 X Pro supports 4K Super Projection, AI fitness, AI motion-controlled games, large-screen karaoke, MeeTime video calls, remote assistance, remote home monitoring, and a dedicated children’s mode. Its remote control supports dual-device operation, allowing users to control both the TV and a set-top box with a single remote. Source

Apple and Google Jointly Launch the Open-Source Eclipsa Video HDR Standard

On June 4, Apple, Google, NBCUniversal, and the Society of Motion Picture and Television Engineers (SMPTE) announced a new HDR video standard called Eclipsa Video. The formal specification is named SMPTE ST 2094-50 and will be managed by the HDR10+ Technologies consortium.

Within the broader HDR standards ecosystem, SMPTE ST 2094-50 is the newest member of the SMPTE dynamic HDR family and is designed to address dynamic HDR adaptation across different display devices. The standard introduces two new forms of dynamic metadata to help displays optimize image rendering based on their individual brightness capabilities.

The first technology is known as the “Reference White Anchor Point.” This mechanism establishes a unified brightness reference for display devices, allowing bright areas in standard dynamic range (SDR) content to be mapped to a fixed reference point while reserving additional brightness headroom for HDR content.

The second technology is called the “Dynamic Luminance Margin Curve.” Through this mechanism, content creators can embed display adaptation instructions directly into video files. When a display has limited brightness capabilities, the content can guide the device to dynamically adjust shadow and mid-tone regions while preserving highlight detail, reducing clipping and loss of image information.

According to the HDR10+ Alliance, Eclipsa Video will first target smartphones before gradually expanding to additional device categories. The first Eclipsa Video-certified products are expected to launch later this year. Because Apple participated in the development of the standard, some industry observers speculate that future iPhone models may be among the first devices to support it, though no official confirmation has been provided. Source

Sony Announces Development of the RIALTO 65 Image Sensor Module

On June 3, Sony announced the development of the RIALTO 65 image sensor module. Designed specifically for Sony’s flagship CineAltaV 2 cinema camera, the module aims to expand the camera’s imaging capabilities from full-frame to the 65mm large-format category.

According to Sony, the RIALTO 65 features a 65mm-format image sensor with a diagonal measurement of approximately 64.60mm, a width of 53.75mm, a height of 35.83mm, and a 3:2 aspect ratio. The module supports full-sensor 9.6K recording in a 3:2 open-gate format and offers multiple readout modes to accommodate different types of 65mm lenses, including certain professional lenses with smaller image circles.

In terms of operation, the RIALTO 65 can either be mounted directly onto the CineAltaV 2 body or connected remotely via cable. Sony stated that the module is fully compatible with the existing CineAltaV 2 ecosystem, allowing the entire system to be upgraded into a 65mm-format digital cinema platform.

Sony plans to showcase the RIALTO 65 at Cine Gear Expo in the United States on June 5, 2026, and will later present an engineering prototype at BIRTV 2026 in China this August. Source

Insta360 Officially Unveils the Luna Ultra Gimbal Camera Head-Tracking Module

On June 3, Insta360 released a new teaser video for the upcoming Luna Ultra and showcased its first-person head-tracking module. The module uses head-direction detection to synchronize the camera gimbal’s movement with the user’s head orientation in real time.

The module features three-axis gimbal tracking and real-time head-motion detection, enabling an immersive shooting experience where “the camera looks wherever you look.” During recording, the system can precisely maintain the user’s point of focus at the center of the frame. The feature also makes it easier to capture dramatic high-angle and low-angle shots that are difficult to achieve with traditional POV setups, expanding the creative possibilities of first-person filming.

For comfort and usability, the module adopts a lightweight ear-hook design and supports multiple shortcut controls. Users can double-tap the module to quickly recenter the gimbal, while a long press can be used to recalibrate its wearing position and other settings. Source

Microsoft Plans to Use AI to Accelerate Windows 11 ARM Ecosystem Compatibility

On June 4, Microsoft hosted a dedicated session at Build 2026 demonstrating how AI can help developers convert x86 applications into native Windows on Arm software, with the goal of further strengthening the Windows on Arm ecosystem.

According to Adams, Microsoft’s Corporate Vice President of Customer Experience Engineering, the Windows on Arm application ecosystem continues to expand. Native Arm applications now account for 90% of total application usage time among Windows on Arm users. Microsoft believes native Arm software provides greater efficiency and higher performance for users’ core workflows.

However, Microsoft also acknowledged that applications without native Arm support still affect user experience and enterprise adoption. Applications that have not yet been ported continue to rely on the Prism emulation layer on Windows on Arm devices. While the combination of native applications and emulation covers a large software library, it does not provide complete compatibility. Some legacy applications, enterprise software, and specialized tools may still perform poorly or fail to launch entirely.

During the session, Microsoft explained that application migration often involves debuggers, profilers, compilers, and other development tools. AI can assist by handling repetitive code transformations, migration checks, and compatibility validation tasks, reducing the overall effort required for developers to adapt x86 applications to native Windows on Arm versions. Source

Apple Store Developer Billings and Sales Surpassed $1.4 Trillion in 2025

Apple announced that the global App Store ecosystem facilitated more than $1.4 trillion in developer billings and sales during 2025, setting a new all-time record.

Of the billings and sales generated through the App Store ecosystem, more than 90% involved no commission payments to Apple. Since 2019, the size of the App Store ecosystem has nearly tripled. Tim Cook stated that developers remain at the heart of the App Store and that Apple will continue providing tools, technologies, and a trusted platform to help them build for the future.

Breaking down the figures, the largest category was physical goods and services, which accounted for approximately $1.1 trillion in sales. Growth was primarily driven by grocery shopping, food delivery and pickup services, general retail, and travel-related services. Digital goods and services generated approximately $149 billion, largely from gaming, enterprise applications, and video streaming platforms. Developer revenue from in-app advertising reached approximately $151 billion.

Artificial intelligence was also highlighted as a major theme in the report. Apple noted that more than 40 of the top 100 App Store applications in 2025 included consumer-facing AI features, and that these apps significantly outperformed other leading applications in terms of revenue growth. AI capabilities are increasingly being integrated into categories such as health and fitness, photo and video editing, and productivity tools, creating new opportunities for developer innovation.

Apple also stated that developers can access the on-device large language models powering Apple Intelligence through the Foundation Models Framework, enabling AI features while preserving privacy and offline functionality without incurring inference costs. Apple added that its development tools, including Xcode, now incorporate AI-assisted coding capabilities to help developers streamline workflows and accelerate product iteration. Source

News Worth a Quick Look

WinFuture has shared a set of promotional images for Microsoft’s upcoming Surface Pro 12. According to the leaked information, the new device is expected to be announced on June 16 and may feature Qualcomm’s Snapdragon X2 Elite processor. In terms of design, the Surface Pro 12 largely continues the familiar Surface Pro aesthetic, remaining very similar to the previous generation. The device retains relatively thick display bezels and the signature multi-angle rear kickstand. Promotional materials indicate that the new model will feature a bright OLED touchscreen designed to deliver high contrast and vivid colors, although no additional display specifications have been revealed beyond that. The leak also suggests that the Surface Pro 12 will be available in Black, Platinum, and Gold color options. On the hardware side, the top-tier 13-inch version is expected to be powered by the Snapdragon X2 Elite ARM processor featuring 12 Oryon cores and up to 32GB of memory. Storage options will reportedly include up to a 1TB PCIe 4.0 SSD, which users will be able to replace themselves. The primary focus of this generation is expected to be local AI capabilities. According to the leaked materials, the Snapdragon X2 Elite’s NPU can deliver up to 80 TOPS of AI performance, which Microsoft claims is sufficient for on-device Copilot-related AI tasks. Battery life is rated at up to 15.5 hours of local video playback, representing a 10% improvement over the previous generation. Source

OpenAI announced an upgrade to ChatGPT’s memory system. The new system is based on the Dreaming V3 mechanism and is designed to better synthesize long-term context such as user preferences, projects, and constraints, allowing ChatGPT to avoid starting from scratch in every conversation. The upgraded system automatically identifies and summarizes important details, presenting them on a dedicated Memory Summary page. Users can review what ChatGPT has learned about them, add or update personal information, and specify which topics should be brought up under particular circumstances. If users want to understand the source of a specific memory, they can also ask the model directly about it. The new memory system is already rolling out to Plus and Pro users in the United States and will expand to additional countries as well as Free and Go users in the coming weeks. Source