Speechly is solving voice in a completely novel way. Our API is fully streaming to enable a few key concepts that we think are the missing piece in voice. We have put together a number of best practices that are worth considering when developing your application.

They are split into four short “chapters”, with the following content:

Chapter 1: Setting the right Context

• Don’t build a Voice Assistant
• Design for Command & Control
• Give visual guidance on what the user can say
• Use voice ONLY for the tasks it’s good for

Chapter 2: Receiving Commands from the User

• Onboard the user
• Avoid using a wake word
• Use a Push-to-Talk button
• Signal clearly when the microphone button is pushed down

Chapter 3: Giving Feedback to the User

• Use non-interruptive modalities for feedback
• Minimize latency with Spoken Language Understanding
• Steer user’s gaze and minimize visual unrest

Chapter 4: Recovering from Mistakes

• Show the transcript
• Produce results fast, but offer opportunity to correct
• Have an intent for verbal corrections
• Offer alternative ways to complete tasks without Voice