Add voice on a microcontroller without having to code -

Add voice on a microcontroller without having to code

Picovoice, a Canadian startup, has launched what it says is the first platform which lets you add a voice interface on your microcontroller without having to write a single line of code.

Its new Shepherd platform, together with its Picovoice console, streamlines adding voice artificial intelligence (AI) onto microcontrollers, enabling the creation of voice experiences similar to Alexa, running entirely on-device and without requiring internet connectivity. Edge voice interfaces built with Picovoice are private, reliable, zero-latency, and cost-effective, distinguishing them from cloud-based alternatives.

Voice models can be created within a users browser to target onto various platforms (Image: Picovoice)

With small low-power and cost-effective microcontrollers already deployed in billions of devices, the ability to bring voice AI to microcontrollers unlocks numerous potential applications that might otherwise be considered infeasible due to additional engineering costs. This is because the technical complexities and specialized skills needed to train and deploy voice AI intro microcontrollers have limited their wide adoption for speech recognition. Picovoice said that today, only a handful of tech giants have access to this technology.

The company’s Shepherd platform simplifies a process that previously would have taken months of R&D by teams of scientists and engineers; it can now be undertaken by a non-technical individual in under an hour. This significantly reduces risk and time to market. The no-code aspect of Shepherd empowers developers, product owners, and designers to create voice interfaces that run entirely on a power-efficient and bargain-priced microcontroller. No coding or machine learning expertise is required.

Picovoice on ST board
An example of adding a voice interface on an  STM32F469,  a development board featuring an Arm Cortex-M4 microcontroller. (Image: Picovoice)
Picovoice upload_firmware_frame
Once the board is connected using a Mini-USB Type-A cable, the Picovoice firmware can be uploaded (Image: Picovoice)

Customers can create voice models within their browsers instantly using Picovoice Console. Once the models are trained, they can be downloaded and loaded onto a microcontroller using Picovoice Shepherd, without any embedded expertise. The process takes only minutes, end-to-end. The company’s web site enables free sign-up to train voice models tailored for your own use case. Using its documentation, Shepherd can be installed on a desktop computer, and voicecan be added using supported development boards from major electronics retailers.

Design, train and test in web browser

Picovoice Console is the company’s platform for designing, training, and testing voice interfaces instantly on an internet web browser, with no machine learning skills required. A user simply describes what is needed in plain text and export trained models. The models run entirely on-device using the Picovoice SDK on these platforms:

  • Embedded platforms such as Raspberry Pi and BeagleBone
  • Android and iOS
  • Modern web browsers
  • Linux (x86_64), macOS (x86_64), and Windows (x86_64)
  • Microprocessors such as ARM Cortex-A
  • Microcontrollers such as ARM Cortex-M

Picovoice Shepherd is free. Picovoice Console is free for personal use (e.g. students, researchers, and tinkerers) and offers a 30-day free trial for enterprise users. The enterprise accounts start from USD 400 per month, charged per annum. Shepherd supports popular Arm Cortex-M-based microcontrollers from ST and NXP, with additional support on the way.

Related Contents:

For more Embedded, subscribe to Embedded’s weekly email newsletter.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.