One of the overarching themes Google is pushing at I/O this year is the miniaturization of voice modeling data — small enough to put into your phone, but robust enough to match what you're saying and deliver new features such as Live Caption with speed and accuracy. But for this feature to work across social media apps, web browsers, and games, it needs to be able to listen to what those apps are putting out. That's now possible with a new audio playback capture API in the latest beta for Android Q.

The new AudioPlaybackCaptureConfiguration API essentially lets a developer with an app that plays audio give other apps permission to record that audio. Recording apps will be able to take audio streams designated from media, gaming, and other in-app sources — nothing from the device's microphones can be picked up. Outputting apps can further control which streams can be recorded or block any recording by non-system apps or all apps for content rights management purposes.

Given the limited audio I/O situation on most Android devices, patching in-app audio to external points such as broadcast equipment may become easier for streamers, journalists, and, perhaps, meme artists.

Back on the theme of miniaturizing voice modeling files, Google is also making big points at I/O 2019 for Continued Conversation for Google Assistant and efficient voice typing on Gboard.