Building Voice-First Android Apps

on YouTube ↗47m

" summary

Android provides SpeechRecognizer and TextToSpeech while ML Kit and TensorFlow handle on-device models; Dialogflow, Amazon Lex, Porcupine, Snips and PocketSphinx are the concrete options shown. Demos show hot-word detection works locally with Porcupine, Snips and PocketSphinx but offline NLU gets wrong outside lexicon for embedded models, and Lex requires cloud setup and occasionally misrecognizes time slots. App Actions require deep links and an actions.xml mapping and app actions map deep links via actions.xml and currently live behind developer preview limits.

AI-written summary · about

" chapters (7)

1. 0:00Start
2. 5:18What do I need to do?
3. 6:46Two strategies
4. 8:47Voice Interactions on Android
5. 31:51Development
6. 37:39App Actions - Developer Preview
7. 40:35Bug Reporting

" related talks

• 47mBuilding Voice-First Android AppsElaine Dias Batista · RS Conf 2019: Mobile People Day
• 42mThe State of Voice Interactions on AndroidElaine Dias Batista · DevFest Paris 2025
• 43mThe State of Voice Interactions on AndroidElaine Dias Batista · AndroidMakers by droidcon 2025
• 35mHaving a Voice on Android: How hard can it be?Paul-Louis Nech · droidcon London 2019
• 29mAndroid Gets a Voice: Building Real-Time AgentsMark Wickham · DroidKaigi 2026
• 36mVoice Enabling Mobile Apps and Custom Hardware DevicesOgnjen Todic · droidcon San Francisco 2022
• 15mDicio voice assistant — Libre, multilanguage and offline assistant for AndroidFabio Giovanazzi · SFSCON 2025
• 55mConversational power for Android appsAlfredo Morresi · DevFest Nantes 2016