Directional Textual Inversion for Personalized Text-to-Image Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kim, Kunhee, Park, NaHyeon, Hong, Kibeom, Shim, Hyunjung
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915851006902272
author Kim, Kunhee
Park, NaHyeon
Hong, Kibeom
Shim, Hyunjung
author_facet Kim, Kunhee
Park, NaHyeon
Hong, Kibeom
Shim, Hyunjung
contents Textual Inversion (TI) is an efficient approach to text-to-image personalization but often fails on complex prompts. We trace these failures to embedding norm inflation: learned tokens drift to out-of-distribution magnitudes, degrading prompt conditioning in pre-norm Transformers. Empirically, we show semantics are primarily encoded by direction in CLIP token space, while inflated norms harm contextualization; theoretically, we analyze how large magnitudes attenuate positional information and hinder residual updates in pre-norm blocks. We propose Directional Textual Inversion (DTI), which fixes the embedding magnitude to an in-distribution scale and optimizes only direction on the unit hypersphere via Riemannian SGD. We cast direction learning as MAP with a von Mises-Fisher prior, yielding a constant-direction prior gradient that is simple and efficient to incorporate. Across personalization tasks, DTI improves text fidelity over TI and TI-variants while maintaining subject similarity. Crucially, DTI's hyperspherical parameterization enables smooth, semantically coherent interpolation between learned concepts (slerp), a capability that is absent in standard TI. Our findings suggest that direction-only optimization is a robust and scalable path for prompt-faithful personalization. Code is available at https://github.com/kunheek/dti.
format Preprint
id arxiv_https___arxiv_org_abs_2512_13672
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Directional Textual Inversion for Personalized Text-to-Image Generation
Kim, Kunhee
Park, NaHyeon
Hong, Kibeom
Shim, Hyunjung
Machine Learning
Computer Vision and Pattern Recognition
Textual Inversion (TI) is an efficient approach to text-to-image personalization but often fails on complex prompts. We trace these failures to embedding norm inflation: learned tokens drift to out-of-distribution magnitudes, degrading prompt conditioning in pre-norm Transformers. Empirically, we show semantics are primarily encoded by direction in CLIP token space, while inflated norms harm contextualization; theoretically, we analyze how large magnitudes attenuate positional information and hinder residual updates in pre-norm blocks. We propose Directional Textual Inversion (DTI), which fixes the embedding magnitude to an in-distribution scale and optimizes only direction on the unit hypersphere via Riemannian SGD. We cast direction learning as MAP with a von Mises-Fisher prior, yielding a constant-direction prior gradient that is simple and efficient to incorporate. Across personalization tasks, DTI improves text fidelity over TI and TI-variants while maintaining subject similarity. Crucially, DTI's hyperspherical parameterization enables smooth, semantically coherent interpolation between learned concepts (slerp), a capability that is absent in standard TI. Our findings suggest that direction-only optimization is a robust and scalable path for prompt-faithful personalization. Code is available at https://github.com/kunheek/dti.
title Directional Textual Inversion for Personalized Text-to-Image Generation
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.13672