VibeToken: Scaling 1D Image Tokenizers and Autoregressive Models for Dynamic Resolution Generations

TL;DR

VibeToken introduces a resolution-agnostic 1D Transformer image tokenizer enabling efficient, flexible autoregressive image synthesis across arbitrary resolutions with significantly reduced computational costs.

Contribution

The paper presents VibeToken, a novel 1D Transformer-based image tokenizer that generalizes to any resolution, and VibeToken-Gen, an efficient AR generator requiring fewer resources.

Findings

VibeToken-Gen synthesizes 1024x1024 images with only 64 tokens and 3.94 gFID.

VibeToken-Gen maintains constant FLOPs regardless of resolution, unlike fixed-resolution models.

VibeToken achieves state-of-the-art efficiency and performance trade-offs in image synthesis.

Abstract

We introduce an efficient, resolution-agnostic autoregressive (AR) image synthesis approach that generalizes to arbitrary resolutions and aspect ratios, narrowing the gap to diffusion models at scale. At its core is VibeToken, a novel resolution-agnostic 1D Transformer-based image tokenizer that encodes images into a dynamic, user-controllable sequence of 32-256 tokens, achieving a state-of-the-art efficiency and performance trade-off. Building on VibeToken, we present VibeToken-Gen, a class-conditioned AR generator with out-of-the-box support for arbitrary resolutions while requiring significantly fewer compute resources. Notably, VibeToken-Gen synthesizes 1024x1024 images using only 64 tokens and achieves 3.94 gFID; by comparison, a diffusion-based state-of-the-art alternative requires 1,024 tokens and attains 5.87 gFID. In contrast to fixed-resolution AR models such as LlamaGen -- whose inference FLOPs grow quadratically with resolution (11T FLOPs at 1024x1024) -- VibeToken-Gen maintains a constant 179G FLOPs (63.4x efficient) independent of resolution. We hope VibeToken can help unlock the wide adoption of AR visual generative models in production use cases.