NanoFLUX: Distillation-Driven Compression of Large Text-to-Image Generation Models for Mobile Devices

TL;DR

NanoFLUX is a compact, 2.4B parameter text-to-image model distilled from a 17B model, enabling high-quality image generation on mobile devices with significant size reduction and latency improvements.

Contribution

The paper introduces a novel compression pipeline including transformer pruning, token downsampling, and text encoder distillation for efficient on-device text-to-image generation.

Findings

Generates 512x512 images in ~2.5 seconds on mobile devices

Reduces model size from 12B to 2B parameters

Maintains high visual quality despite compression

Abstract

While large-scale text-to-image diffusion models continue to improve in visual quality, their increasing scale has widened the gap between state-of-the-art models and on-device solutions. To address this gap, we introduce NanoFLUX, a 2.4B text-to-image flow-matching model distilled from 17B FLUX.1-Schnell using a progressive compression pipeline designed to preserve generation quality. Our contributions include: (1) A model compression strategy driven by pruning redundant components in the diffusion transformer, reducing its size from 12B to 2B; (2) A ResNet-based token downsampling mechanism that reduces latency by allowing intermediate blocks to operate on lower-resolution tokens while preserving high-resolution processing elsewhere; (3) A novel text encoder distillation approach that leverages visual signals from early layers of the denoiser during sampling. Empirically, NanoFLUX generates 512 x 512 images in approximately 2.5 seconds on mobile devices, demonstrating the feasibility of high-quality on-device text-to-image generation.