A 1.75M-Parameter Decoder for Kokoro: No Neural Network at Audio Rate

Wed, 11 Nov 2026 12:00:00 +0300

The previous post shrank Kokoro-82M’s decoder by pruning and distillation, and ran into a wall: the speed ceiling is set by the generator’s last stage, which runs a neural network 4,800 times a second, and that stage is exactly where the quality lives. This post takes the opposite approach. The decoder does all of its learned work at 160 frames per second, and the audio-rate part is fixed signal processing with nothing to learn.

Edge-Ai on DRM HSE

A 1.75M-Parameter Decoder for Kokoro: No Neural Network at Audio Rate