Picovoice on-device speech-to-text engines slash the requirements and cost of transcription

Speech-to-text benchmarks accuracy

Picovoice Leopard and Cheetah offline, on-device speech-to-text engines are said to achieve cloud-level accuracy, rely on tiny Speech-to-Text models, and slash the cost of automatic transcription by up to 10 times. Leopard is an on-device speech-to-text engine, while Cheetah is an on-device streaming speech-to-text engine, and both are cross-platform with support for Linux x86_64, macOS (x86_64, arm64), Windows x86_64, Android, iOS, Raspberry Pi 3/4, and NVIDIA Jetson Nano. Looking at the cost is always tricky since companies have different pricing structures, and the table above basically shows the best scenario, where Picovoice is 6 to 20 times more cost-effective than solutions from Microsoft Azure or Google STT. Picovoice Leopard/Cheetah is free for the first 100 hours, and customers can pay a monthly $999 fee for up to 10,000 hours hence the $0.1 per hour cost with PicoVoice. If you were to use only 1000 hours out of your plan that […]

Intel to invest $1 billion in foundry innovation, becomes RISC-V International member

Intel RISC-V

Intel has just announced a $1 billion fund to support companies bringing innovations and new technologies to the foundry ecosystem. The company says the fund will prioritize investments in “capabilities that accelerate foundry customers’ time to market – spanning intellectual property (IP), software tools, innovative chip architectures, and advanced packaging technologies.” What’s interesting is that it does not only cover x86 architecture but also Arm and RISC-V, with a focus on the latter, as Intel has just become a Premier member of RISC-V International, and partnered with several companies offering RISC-V solutions including Andes Technology, Esperanto Technologies, SiFive, and Ventana Micro Systems. Intel’s Open Chiplet Platform Part of the investment will go to the Open Chiplet Platform offering a modular approach to chip design through chiplets with each block/chiplet customized for a particular function. This will allow designers to select the best IP and process technologies for a particular SoC. […]

Tiny Core Linux 13.0 released for older or lower-end x86 hardware

Tiny Core Linux 13

Tiny Core Linux 13.0 has just been released for 32-bit and 64-bit x86 systems with a recent Linux 5.15.10 kernel, and various upgrades to the packages. Tiny Core Linux is a lightweight (~22MB ISO) Linux distribution with an FLTK (Fast Light Toolkit)/FLWM (Fast Light Windows Manager) desktop and based on the Core Project that integrates a recent Linux kernel, vmlinuz, and a root filesystem with low footprint libraries such as busybox. It’s mostly interesting for older or low-end hardware that may be slow and/or unusable with more common Linux distributions like Ubuntu or Debian. Tiny Core Linux 13.0 main changes: kernel updated to 5.15.10 glibc updated to 2.34 gcc updated to 11.2.0 binutils updated to 2.37 e2fsprogs base libs/apps updated to 1.46.4 util-linux base libs/apps updated to 2.37.2 busybox updated to 1.34.1 50-udev-default.rules: add media permissions select: require that “break” used in awk scripts filetool.sh: prevent gratuitous change to .filetool.lst’s […]

Year 2021 in review – Top 10 posts and statistics

cnx software happy new year 2022

As per tradition, we’ll look back at what happened during the year in the last post, and see what 2022 may have in store, plus the usual statistics from CNX Software website. The biggest story of 2021 has to be the worsening of semiconductors shortages with extremely long lead times, prices of some components going up multiple folds, constant complaints on Twitter about availability and prices. I think I even saw a website, hopefully misconfigured, showing an estimated availability of a specific STM32 MCU in 2037. This also gave rise to opportunities and board redesigns, with MotorComm Ethernet chips replacing some Realtek chips in SBCs such as NanoPi R2C and  Orange Pi R1S Plus LTS, and CH9102F showing up as a replacement for CP2104 in some IoT boards. We also got some interesting Arm processors, but sadly the high-expected Rockchip RK3588 got delayed by another year, although it’s getting really […]

BrainChip AKD1000 SNN AI SoC gets Raspberry Pi and x86 development kits

Brainchip Akida Raspberry Pi & Mini PC devkits

BrainChip has introduced two development kits for its Akida AKD1000 neuromorphic processor based on Raspberry Pi and an Intel (x86) mini PC in order to enable partners, large enterprises, and OEMs to begin testing and validation of the Akida chip. BrainChip Akida neural relies on spiking neural networks (SNN) which enable high-performance, real-time inference at ultra-low power, notably much lower power than traditional AI chips relying on CNN (convolutional neural network) technology. Akida Development Kit based on Raspberry Pi CM4 Specifications: SoM – Raspberry Pi CM4 or CM4 Lite with SoC: Broadcom BCM2711C0 quad-core ARM Cortex-A72 (ARMv8-A) 64-bit @ 1.5GHz plus Broadcom VideoCore VI GPU RAM – 1GB, 2GB, 4GB, or 8GB LPDDR4 SDRAM Storage – MicroSD card for CM4 Lite, or 2GB to 32GB eMMC for CM4 Networking – Optional 2.4 GHz and 5 GHz 802.11b/g/n/ac Wi-Fi, Bluetooth 5.0 LE, Gigabit Ethernet PHY Carrier board –  Official Raspberry Pi […]

ODROID-H2+ SBC discontinued due to supply shortage

ODROID-H2+ discontinued

Hardkernel has just discontinued ODROID-H2+ single board computer based on the Intel Celeron J4115 Gemini Lake Refresh processor, which followed ODROID-H2 SBC itself being discontinued shortly after Intel decided to phase out Intel J4105 and other Gemini Lake processors. The reason given is the “uncertain situation of main component supply”, which could mean Celeron J4115 processor is hard to get (or expensive), or the Realtek RTL8125B chipset provides 2.5GbE networking. That means Hardkernel does not offer any x86 SBC at this time. That’s a shame before ODROID-H2+ was a well-supported SBC running Linux or Windows, and great value for money at $119, especially for people interested in the two 2.5 Gbps Ethernet ports found on the board (and upgradeable to six), not to mention support for SO-DIMM memory and M.2 NVMe SSD. Update: There were multiple components with very long lead times that must have contributed to the situation as […]

Overview and List of System-on-Module and Computer-on-Module Standards – Q7, SMARC, COM HPC, and More

QSeven_vs_SMARC

A System-on-Module (SoM), also known as a Computer-on-Module (CoM), is a small board with the key components of a computer such as SoC, memory, and possibly others components such as PMIC (Power Management IC), an Ethernet PHY, as well as one or more connectors used to connect to a baseboard, also called carrier board, which features standard ports such as Ethernet (RJ45), USB ports, SATA, power jack and so on. The advantages of using of baseboard + SoM design compared to a single board are at least twofold: Most of the PCB design complexity is often around the CPU/SoC and high-speed buses connected to the CPU/SoC. So you could buy an SoM, design your own baseboard and get a complete design relatively in a short amount of time, with reduced development resources and costs. The design is modular, so you could easily upgrade from one SoM to another one. For […]

QNAP TVS-675 NAS features Zhaoxin KaiXian KX-U6580 x86 processor

QNAP TVS-675 NAS

QNAP has been trying something different in several in their new NAS recently. Last month, we covered QMiroPlus-201W that combines an Arm router and Intel NAS into one device, and now the company has announced another x86 NAS, but not based on Intel or AMD processor, but instead Zhaoxin’s 2.5 GHz eight-core KaiXian KX-U6580 processor. QNAP TVS-675 NAS comes with 8GB RAM by default, six drive bays for hard drives or SSD’s, two M.2 2280 slots for SATA or NVMe storage, as well as two PCIe 3.0 x4 slots for networking, storage, or other expansion cards. The NAS also features 2.5GbE wired connectivity and an HDMI 2.0 to connect a display. QNAP TVS-675 specifications: SoC – Zhaoxin KX-U6580 8-core/-thread 64-bit x86 processor @ 2.5GHz with FPU, AES-NI accelerator, 8MB L2 cache, ZhaoXin C-960 Graphics; 16nm process System Memory – 8 GB DDR4 single-channel; upgradable to up to 64GB dual-channel memory […]

EmbeddedTS embedded systems design