Saved in:
Bibliographic Details
Main Authors: Beyer, L. Lao, Li, T., Chen, X., Karaman, S., He, K.
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.08257
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912422336397312
author Beyer, L. Lao
Li, T.
Chen, X.
Karaman, S.
He, K.
author_facet Beyer, L. Lao
Li, T.
Chen, X.
Karaman, S.
He, K.
contents Commonly used image tokenizers produce a 2D grid of spatially arranged tokens. In contrast, so-called 1D image tokenizers represent images as highly compressed one-dimensional sequences of as few as 32 discrete tokens. We find that the high degree of compression achieved by a 1D tokenizer with vector quantization enables image editing and generative capabilities through heuristic manipulation of tokens, demonstrating that even very crude manipulations -- such as copying and replacing tokens between latent representations of images -- enable fine-grained image editing by transferring appearance and semantic attributes. Motivated by the expressivity of the 1D tokenizer's latent space, we construct an image generation pipeline leveraging gradient-based test-time optimization of tokens with plug-and-play loss functions such as reconstruction or CLIP similarity. Our approach is demonstrated for inpainting and text-guided image editing use cases, and can generate diverse and realistic samples without requiring training of any generative model.
format Preprint
id arxiv_https___arxiv_org_abs_2506_08257
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Highly Compressed Tokenizer Can Generate Without Training
Beyer, L. Lao
Li, T.
Chen, X.
Karaman, S.
He, K.
Computer Vision and Pattern Recognition
Artificial Intelligence
Commonly used image tokenizers produce a 2D grid of spatially arranged tokens. In contrast, so-called 1D image tokenizers represent images as highly compressed one-dimensional sequences of as few as 32 discrete tokens. We find that the high degree of compression achieved by a 1D tokenizer with vector quantization enables image editing and generative capabilities through heuristic manipulation of tokens, demonstrating that even very crude manipulations -- such as copying and replacing tokens between latent representations of images -- enable fine-grained image editing by transferring appearance and semantic attributes. Motivated by the expressivity of the 1D tokenizer's latent space, we construct an image generation pipeline leveraging gradient-based test-time optimization of tokens with plug-and-play loss functions such as reconstruction or CLIP similarity. Our approach is demonstrated for inpainting and text-guided image editing use cases, and can generate diverse and realistic samples without requiring training of any generative model.
title Highly Compressed Tokenizer Can Generate Without Training
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2506.08257