Finding Your Voice: Building Screenless Interfaces with Node.js
summary
NPR builds a Node.js voice stack that uses AWS Lambda (or any REST endpoint) and official SDKs to handle Alexa and Google Assistant requests, sharing about 60% shared code via per-platform BFFs and a generic response model. TypeScript powers controllers while views use SDK helpers to send speech or audio. SSML fixes handle acronyms like SSML say-as characters. Google’s native audio player lags, and the real bottleneck is manual QA—no reliable end-to-end voice test framework exists yet.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:20What is NPR?
- 1:13Why voice UI development?
- 5:57A brief timeline of voice assistants
- 9:54Conclusions
- 11:57The request/response flow
- 15:44The official SDKs are not bad
- 15:58Examples from Alexa SDK
- 17:05SSML: A common language
- 21:00Generic Response Model
- 21:42Challenges
- 25:49Open source opportunities
related talks
26mFinding Your Voice: Building Screenless Interfaces with Node.js
29mBuild and scale multiple Voice application by using TypeScript
21mBuilding a Voice-Enabled AI Assistant With Javascript
44mDeveloping Skills & Actions for Voice Assistants
48mConversational ULS: Talking to Siri, Alexa, and Your Web Browser
19m"Bringing Voice First Experiences to the Web"
49mDesigning voice interfaces
38mConversational UIs: Talking to Siri, Alexa, and Your Web Browser