Privacy by Design is an approach to systems engineering which takes privacy into account throughout the whole engineering process. This principle will be mandatory for every company offering a product or service to a European citizen after May 2018.
Amazon Alexa, Google Assistant, Dialogflow, Microsoft Cortana, Luis.ai, Facebook Wit.ai and most solutions out there run in the cloud: If you power your assistant with one of these, your voice or the one if your users will be systematically sent to the cloud. Such solutions are not Private by Design. They cannot run offline. And if you are building a product out of them, it will cost you every time someone uses your product.
Snips, on the other hand, runs completely on device, with nothing being sent to the cloud. This means it guarantees your Privacy, works offline and doesn't have variable costs!
Snips is completely free for makers and for building prototypes. For commercial use (products you sell), we charge fixed fees per device, for unlimited queries. You can learn more on our enterprise page.
Today, the best way to help us is to provide feedback. Join our community on our Forum. Stay tuned though, because we will soon open source the platform!
Snips' Natural Language Understanding understands English, French, Spanish, German, Japanese and Korean.
As for Speech Recognition, it currently understands English, French, German and Japanese, but more languages will be added soon!
If you need ASR in another language than ones supported by snips, we provide an integration with Google's Cloud Speech service.
Snips offers two types of Wakewords: Universal and Personal ones.
Universal Wakewords are meant to be triggered by anyone saying the Wakeword. They require models that have been trained on a wide variety of voices. On the Snips console, you can pick a Universal Wakeword out of a list we will be progressively extending. Creating custom Universal Wakewords is part of our enterprise offering. Reach out if you're interested.
Personal Wakewords can be arbitrary, you simply need to provide a few samples following this tutorial. Note that these Wakewords Detectors are based on a much simpler logic, and may not necessarily work for other voices than the one who provided the training samples.
Debian — Available via apt standard packaging for armhf architecture.
Raspbian — Available via apt standard packaging on Stretch.
OS X — Support coming soon via Brew.
Yocto — Please contact us for support specs.
Android — Available as a SDK you can add to your application. We support api 21+ on armv7, aarch64, x86 and x86_64.
iOS — Available as a SDK for iOS11 and later. You can use it in your application.
Windows — Coming soon.