You can experiment with buffer settings. https://github.com/EpicGames/UnrealE…pture.cpp#L148 And look into capture frequency as well to make sure you grab the packets in time.
https://docs.unrealengine.com/portals/0/images/Support/Builds/ReleaseNotes/4_18/image_9.jpg
Generally speaking, lower the buffer size the lower the latency will becomes, but this is only a layer for directsound which is never going to produce low latencies for you. That’d require a different audio capture / rendering method, and in many cases special hardware, such as a studio grade sound card, stuffed directly into a shiny PCI-E slot.
Audio is rather similar to video playback, as it always require high performance to keep it in sync with whatever medium or event triggered it. The problems for video is solved by GPU / graphics accelerator, that is a reliable hardware to produce this content in close to realtime. But what about the audio? This is a key issue that no audio accelerators available in consumer pcs - not really - the one you call sound card integrated on your motherboard is a very cheap quality equipment not designed for low latency works.
But of course you can have an expensive sound card in your computer, and use advanced audio rendering techniques, such as ASIO, WASAPI, Kernel Streaming to allow low latency playback /o record. These however start to point out of the context of gaming, since the CPU costs of such high performant use cases can easily get in the way of gameplay, and don’t underestimate the compatibility issues with average user rigs.
Directsound should be considered the worse of all - in the context of record / playback latencies - however it is the most reliable, highest compatibility and cpu cost efficient method to produce audio which is in favor of many game specific reasons. And it is duplex (parallel in/out) that will easily choose playback over recording when it comes for sharing the resources. No luck here with low latency recording.
The RtAudio lib this component is based on - actually packed with many different playback solutions as well, it just not implemented in the engine as of yet. GitHub - thestk/rtaudio: A set of C++ classes that provide a common API for realtime audio input/output across Linux (native ALSA, JACK, PulseAudio and OSS), Macintosh OS X (CoreAudio and JACK), and Windows (DirectSound, ASIO, and WASAPI) operating systems.