What is WASAPI? PC audio for live captions on Windows
WASAPI is the Windows Audio Session API. It lets applications work with audio from a device. Its loopback mode captures the mix played through a PC output; it does not require a microphone listening to the speakers. To caption speech from a video or call, first choose that output or your microphone according to whose words you need to read.
This guide separates audio routing from transcription. WASAPI carries audio: the receiving application performs speech recognition and translation. It is neither a language pack nor a guarantee of correct captions.
System audio or microphone: choose the voice
- The other person speaks through a PC application: start with system audio. Identify the output where you hear that voice, such as your headphones.
- You want to transcribe your own voice: you need microphone input. Hearing a video does not establish that your microphone works.
- You need both sides: Aurora documents Both, an experimental mode with separate inputs. Microphone text appears in Session Activity, not the main overlay. Its absence from the overlay does not establish that capture failed.
Loopback on an output can capture its mix of sounds. Do not confuse this with automatic isolation of one application or browser tab. For the first check, play one speech source and pause other content.
A controlled check in Aurora
Choose a speech clip you have permission to play and whose words you can check. This is a proposed procedure for your PC, not an Aurora accuracy or latency measurement.
- Check the starting conditions. Aurora requires 64-bit Windows 11. Read Aurora’s current requirements and modes before downloading models or choosing settings.
- Listen to the source. Play the clip and check which output you hear it through. If you switch between speakers and headphones, check the route again before continuing.
- Choose system audio in Aurora for speech played by the PC. Keep one source for this check; do not use your own speech to test that route.
- Check the source language and selected mode. Maximum Compatibility accepts English audio; for another language, check the Maximum Quality options. The preset and language guide expands on that choice.
- Compare the text with the clip. Check whether the words appear and inspect omissions. Then adjust overlay readability. Add translation in a later check so that you can distinguish the two issues.
I can hear audio, but no captions appear
Separate what you have observed from what you have yet to check:
- You cannot hear the source: check its playback, volume and Windows output first. You have not yet established Aurora’s input.
- You hear speech but Aurora shows no text: check the selected source and the language supported by the mode again. Audible playback alone does not establish that the application is capturing it.
- Text comes from other content: pause the other sources and repeat the check. Mixed output is not an isolated conversation.
- Your voice is missing in Both: check Session Activity and microphone input; the overlay is not the documented destination for microphone text.
Microsoft documents two technical limits of classic loopback: it uses a shared-mode stream, and trusted drivers can prevent capture of protected content. Do not assume that every player or piece of content can be captured or try to bypass its protection.
Choose your next step
If you only need Windows captions, also check Microsoft’s built-in feature. If you are considering Aurora, compare its inputs, languages and overlay on the product page before choosing. This check helps define your use case; it does not guarantee perfect recognition or replace review of important content.