Pull a voice out of noise.
Drag the focus ring. Cut the steady bed without sanding away consonants, then hear what your choice actually did.
Why this belongs here. Huawei describes Pure Voice Calling as a network-level AI sequence: Multi-dimensional Feature Clustering identifies noise, then HiDCRNN restores voice detail. This room makes that separation/restoration tension audible. It uses deterministic local DSP, not Huawei's proprietary models.