Channel-wise Vector Quantization

We present Channel-wise Vector Quantization (CVQ), a novel image tokenization paradigm that replaces patch-wise tokens with channel-wise tokens. Unlike conventional vector quantization, which assigns a discrete token to each patch feature vector, CVQ quantizes each channel of the feature map. This formulation represents an image as discrete levels of visual details, rather than as a grid of spatial patches. Based on CVQ, we introduce a new visual autoregressive framework with "next-channel prediction". Instead of rendering images patch by patch in raster order, our Channel-wise Autoregressive (CAR) model predicts image channels sequentially, producing progressively enriched visual details. Specifically, it first sketches global structure and then refines fine-grained attributes, akin to a human artist's workflow. Empirically, we show that: (1) CVQ achieves 100% codebook utilization with a 16K+ codebook size without any bells and whistles, while reducing reconstruction FID by 50% over conventional VQ; and (2) CAR outperforms the AR baseline by improving the GenEval score from 0.69 to 0.74 and the DPG score from 79.86 to 82.14, demonstrating strong effectiveness for text-to-image generation. We hope our research offers a new perspective on the fundamental unit of visual tokenization by moving from spatial patches to channels.