# Low-data training

Soundwave is a groundbreaking speech-to-text model that bridges the gap between speech and text, demonstrating exceptional performance in speech translation and AIR-Bench speech tasks with just 10,000 hours of training data.

ControlNet v1.1 is a neural network structure that controls pre-trained large diffusion models by adding additional conditions. It is particularly suitable for image generation and super-resolution tasks based on tile image conditions.