The French Startup Snips now helps you create a custom voice assistant for your device. Snips do not use Alexa Voice Service or Google Assistant SDK from Amazon – the company builds its own voice assistant so you can integrate it on your devices. And the best part is that it does not send anything into the cloud because it works offline.
If you want to understand how a voice assistant works, you can divide it into parts. First, it starts with a password. Snips has a handful of default standby words, such as "Hey Snips", but you can also pay the company to create your own password.
For example, if you build a multimedia robot called Keecker, you can create a special word "Hey Keecker". Snips then uses in-depth learning to accurately detect when someone is trying to speak to your voice assistant.
The second part is automatic speech recognition. A voice assistant transcribes your voice into a text query. Popular family assistants usually send a small audio file with your voice and use servers to transcribe your request.
The shears can transcribe your voice into text on the device itself. It works on anything that is more powerful than a Raspberry Pi. For now, Snips is limited to English and French. You will need to use a third-party automatic voice recognition API for other languages.
Next, Snips must understand your request. The company has developed natural language skills. But there are hundreds or even thousands of ways to ask a simple question about the weather, for example.
That's why Snips today launches a data generation service. I saw a demo yesterday, and the interface looks like Automator on macOS or Workflow on iOS. You define certain variables, such as "date" and "location", you define if they are mandatory for the query and you enter some examples.
But instead of manually entering hundreds of variations of the same query, you can pay from $ 100 to $ 800 to allow Snips to do the work for you. Startup manually checks your request, then publishes it on Amazon Mechanical Turk and other crowdsourcing markets. Finally, Snips cleans your dataset and sends it back to you.
You can download it and reuse it in another chatbot or voice assistant, or you can use it with the Snips voice wizard. You can also make your ability public. Other Snips users can add this feature to their own wizard by browsing a repository of pre-trained capabilities.
Step 1. Create the Intention
Step 2. Choose the datagen package
Step 3. Confirm the results
A Snips voice wizard usually requires hundreds of megabytes, but it's pretty easy to update. After installing the Snips application on your device, you just need to replace a zip library file to add new features.
You must also implement the actual actions. Snips only translates what someone says in an analysable query. For example, Snips can understand that "could you turn on the light in the bedroom?" Means "light + bedroom + on". A developer must always implement action based on these three parameters.
Developers already play with Snips to test their abilities. But the company hopes that major device manufacturers will integrate Snips into their future products. Finally, you might think of a coffee machine with a voice assistant Snips.
Compared to the large-scale assistants of Amazon or Google, Snips thinks you do not need to integrate a complete voice wizard into all your devices. You just want to tell your Roomba to start vacuuming – no need to let you start a Spotify playlist of your vacuum cleaner.
This approach has some advantages in terms of confidentiality and network effects. Large technology companies are creating an ecosystem of Internet-of-things devices. People buy light bulbs, security cameras and door locks that work with the Amazon Echo for example.
But if you can talk to the devices themselves, you do not need to plug in your devices with a center speaker – the central hub disappears. If voice assistants are more than fashionable, Snips builds a promising technology. And Snips could get license revenue for each device provided with its voice assistant.
Stock Image: Bryce Durbin / TechCrunch