(Update: Snowboy had more than 3000 developers by 10/29/2016).

As a developer, we knew the exact pains every one of us had been facing before the era of Snowboy when searching solutions for always-listening voice activation:

  1. online solutions (from Google/Bing/Watson/Nuance) are costly and do not protect privacy
  2. offline solutions (using a small speech recognition engine) are very resource intensive and do not fit on embedded devices.

Thus we developed Snowboy, the super-fast and super-lean offline hotword detection engine powered by our in-house Deep Neural Network:

In May 2016 we launched Snowboy. Now it has more than 1,000 developer users and the hotword library contains more than 500 unique hotwords!

Who is using Snowboy

Snowboy is serving both the enterprise customers and the developer community. As we have envisioned, the developer community quickly picked up Snowboy whenever there is a need for voice activation. I'm listing a few popular GitHub projects that are actively using Snowboy.

  • Smart Mirror (yes, a talking and listening mirror!)

  • Jarvis (you know this one, J.A.R.V.I.S.)

  • AlexaPi (running Amazon's Alexa Voice Service on a Raspberry Pi)

Snowboy makes it much easier for developers to prototype their voice enabled projects. And we've seen an increasing number of users from fields related to home automation, robots, human machine interaction, etc.

What developers say about Snowboy

Snowboy has received quite positive feedbacks from developers! Developers had been looking for solutions like Snowboy for a long time before its official release in May 2016, but often ended up with lots of painful dead ends. I'm citing a few comments we collected from developers (with permissions) that reflect the excitement from the developer community:

Snowboy Evan Cohen Comment

  • It's worlds apart for hotword detection. (Jasper developer)

Snowboy Evan Cohen Comment

  • ultra rapide (yes, it's from a French Jarvis developer!)

Snowboy alexylem Comment

My favorite part comes from the wiki page of alexylem's jarvis project, where they compared Snowboy with various speech-to-text engines and concluded that Snowboy is the best for keyword/hotword detection. The original page is in French, so I used Google Translate and took a screenshot for the English version:

Snowboy Performance Comparison

Demos by developers

Developers have been posting demo videos of their projects using Snowboy. Here are a few!

  • Snowboy + Jasper on Raspberry Pi (credit to Lee Melbourne)
  • ADRIAN -- open source virtual assistant for the Raspberry Pi (credit to the ADRIAN team)

Where to go from here

KITT.AI is dedicated to making difficult and challenging tasks easy and intuitive to both developers and end consumers. We will be continuing supporting Snowboy, its hotword library and the whole developer community. Together we bring the future of human-machine interaction to the present.

Developers, share your awesome demos with us and we'll update this blog post accordingly. Companies who are interested in commercializing can make an inquiry to snowboy@kitt.ai.