# rnnoise-models

This repository contains the tools and information used to create the various
neural network models in rnnoise-nu. With the exception of the tools/ directory
and this file, none of this work is creative and thus none of it is subject to
copyright.

Note that the name of each model is not meaningful.

## Model information

The following table maps expected signal (horizontal) against expected noise
(vertical). Note that "voice" is distinct from "speech" by the presence of
non-speech human sounds, such as laughter.

|                   | General               | Voice                 | Speech                |
| ----------------- | --------------------- | --------------------- | --------------------- |
| General           | marathon-prescription | leavened-quisling     | orig                  |
| Recording         | conjoined-burgers     | beguiling-drafter     | somnolent-hogwash     |