RT-Thread has launched a cloud supply campaign for Edgi-Talk, a handheld reference platform built for hardware-accelerated machine learning. It is based on Infineon's PSoC Edge E84 architecture and targets smart home, wearable, and industrial interfaces that rely on local inference and responsive voice interaction.
The core of the system is the Infineon PSoC Edge E84, which combines a 400 MHz Arm Cortex-M55 processor with Helium DSP enhancements and a 200 MHz Cortex-M33 core. Machine learning workloads are supported by Arm Ethos-U55 micro NPUs and Infineon's ultra-low-power NNLite accelerators.
Edgi-Talk peripherals
Edgi-Talk integrates a 4.3-inch MIPI-DSI capacitive touchscreen within a 115 x 73 x 20 mm handheld housing. Wireless connectivity is provided by the Infineon AIROC CYW55512 module and supports dual-band Wi-Fi 6 and Bluetooth 6.0.
The board includes 128 MB of PSRAM, 128 MB of QSPI Flash, dual digital microphones with 1 W amplifier, AHT20 environmental sensor, and LSM6DS3TR-C 6-axis IMU.
Expansion options include a Raspberry Pi-compatible 40-pin header, PMOD interface, USB-HS, MicroSD slot, and standard I²C, UART, and SPI interfaces.
Installing the Edgi-Talk battery
The platform's GitHub repository shows that it runs the RT-Thread OS and includes the XiaoZhi language model to demonstrate offline interaction capabilities. RT-Thread provides a fully open-source board support package that includes firmware, drivers, schematics, and sample applications for use with RT-Thread Studio.
Power is provided via a 5V USB Type-C port and supports an internal 500mAh 502540 Li-ion battery that can be installed separately.
edge talk
Detailed information
The Edgi-Talk campaign starts this week and costs $139. As of this writing, we have raised $139 towards our goal of $13,900. Manufacturing takes place in China, and fulfillment is handled by Mouser Electronics through a cloud supply logistics pipeline.
The research team notes that while the manufacturing plan takes into account current pricing conditions as of December 2025, future pricing changes could impact costs. In a worst-case scenario, it means customers may be required to refund money or pay additional fees.
