Wake Word

  • State of the art deep learning
  • Pick your own wake word
  • Data collection and model training as a service
  • Multiple wake words gateway

Hey Snips, tell me the weather

Speech to Meaning

  • State of the art deep learning
  • 100% on device ASR & NLU
  • Real time speech recognition
  • NLU runs in <100ms
  • English, French, German and Japanese supported end-to-end. More coming soon
  • Data Generation solution fixes cold start problem
  • Specialised for your vocabulary

Hey Snips, could you order another EKTORP armchair from Ikea, please?

Hey Snips, I’m cold, would you mind closing the window, please?

Dialog

  • Built-in support of missing fields in user queries
  • Multi-room support: let multiple people simultaneously interact with your voice assistant at home
  • Multi-turn: build complex interactions, beyond simple command-execution commands

Book me a flight from Paris to New York

No problem, do you prefer to fly from Orly, or from Charles de Gaulle?

Data Generation

Save months of work by using our Human-in-the-loop algorithm to generate training examples

  • Describe
    your intents

    Create custom intents and describe them by giving a few examples.

  • Generate
    data for it

    We will generate thousands of training examples using a patented combination of human operators and algorithms.

  • Train
    your assistant

    You can then use the data in Snips, or download to use on other platforms.

Performances

“Can you dim a bit the lights in the kid's bedroom?”
“Could you switch the garden lights to blue?”
93 %
Sound to meaning F1-score
More info

Our Current Research

This paper presents the machine learning architecture of the Snips Voice Platform,
a software solution to perform Spoken Language Understanding on microprocessors typical of IoT devices.
Read more

Embedded Spoken Language Understanding

Today, inference runs fully on device and training is still centralized on our servers on data we generated without using any user data. The next frontier is to be able to train our models directly on real user data, in a private fashion.

Read more

Federated learning at Snips - towards a private AI

Today, we are open sourcing Snips NLU. It can run on the Edge or on a server, with minimal footprint, while performing as good or better than cloud solutions.

Read more

Open Sourcing Snips NLU

We open-sourced a set of useful scripts to simplify the TensorFlow cross-building process. We made them all available on github.

Read more

How we made Tensorflow run on a Raspberry Pi

The wake word is used to start a conversation with a voice assistant. We take this example as an illustration of how Machine Learning is done on voice.

Read more

Machine Learning on Voice: a gentle introduction with Snips Personal Wake Word Detector

We recently open-sourced a benchmark datasets of over 16K queries on which we benchmarked the main actors of the field: Google, Facebook, Microsoft, and Snips.

Benchmarking NLU systems

Ready to build your voice assistant?

Start Building for Free
Group 10Group 6androidD698AA58-CB05-4BE4-B9CF-4AED4D91624F4FC2F125-2C2F-4727-BB99-0DBE71289E0A965BFF08-315B-48F5-B813-291F02983210appleautomotivebankingbatteryic_event_black_24px(2)4504010F-A4E7-489E-8532-DB84DBC3BB72integration_bluesetupNLQGroup 23domain_customR4.1Artboard 65 copy 2Shapeentertainment3DA8B603-BE12-4AB6-868B-617C22B72D15ic_home_black_24px(2)integrationinternationaliot copy 393C6650D-D3AA-4FD6-BB85-718D4204A2BDic_room_black_24px(1) copydevicelodgingGroup 5BETAGroup 20Group 18reservations copyofflineKG0B180F11-DA95-41A2-A187-B1D428B0E4A1placesofflineF45EAB3D-2F56-4866-892B-DFB3DDECFB76reservationsR4.1Artboard 65retailrocket_greensetupArtboardsmart homeD06202CF-344D-4FD5-A717-5DB5BD84880Bsupport_customtransitusers_blueweather