An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion (2022-08-02T00:00:00.000000Z)

TL;DR

This work uses only 3-5 images of a user-provided concept to represent it through new words in the embedding space of a frozen text-to-image model, and finds evidence that a single word embedding is sufficient for capturing unique and varied concepts.

Abstract

Text-to-image models offer unprecedented freedom to guide creation through natural language. Yet, it is unclear how such freedom can be exercised to generate images of specific unique concepts, modify their appearance, or compose them in new roles and novel scenes. In other words, we ask: how can we use language-guided models to turn our cat into a painting, or imagine a new product based on our favorite toy? Here we present a simple approach that allows such creative freedom. Using only 3-5 images of a user-provided concept, like an object or a style, we learn to represent it through new"words"in the embedding space of a frozen text-to-image model. These"words"can be composed into natural language sentences, guiding personalized creation in an intuitive way. Notably, we find evidence that a single word embedding is sufficient for capturing unique and varied concepts. We compare our approach to a wide range of baselines, and demonstrate that it can more faithfully portray the concepts across a range of applications and tasks. Our code, data and new words will be available at: https://textual-inversion.github.io

Authors

Gal Chechik

11 papers

Rinon Gal

4 papers

Y. Atzmon

4 papers

TL;DR

Abstract

Authors

References93 items

Imagic: Text-Based Real Image Editing with Diffusion Models

DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation

Classifier-Free Diffusion Guidance

Don't Stop Learning: Towards Continual Learning for the CLIP Model

Authentic volumetric avatars from a phone scan

Scaling Autoregressive Models for Content-Rich Text-to-Image Generation

Blended Latent Diffusion

On Aliased Resizing and Surprising Subtleties in GAN Evaluation

Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding

TEMOS: Generating diverse human motions from textual descriptions

VQGAN-CLIP: Open Domain Image Generation and Editing with Natural Language Guidance

Hierarchical Text-Conditional Image Generation with CLIP Latents

No Token Left Behind: Explainability-Aided Image Classification and Generation

Text2LIVE: Text-Driven Layered Image and Video Editing

"This is my unicorn, Fluffy": Personalizing frozen vision-language representations

MotionCLIP: Exposing Human Motion Generation to CLIP Space

Name Your Style: An Arbitrary Artist-aware Image Style Transfer

Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution

High-Resolution Image Synthesis with Latent Diffusion Models

GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models

CLIP2StyleGAN: Unsupervised Extraction of StyleGAN Edit Directions

Text2Mesh: Text-Driven Neural Stylization for Meshes

CLIPstyler: Image Style Transfer with a Single Text Condition

HyperInverter: Improving StyleGAN Inversion via Hypernetwork

HyperStyle: StyleGAN Inversion with HyperNetworks for Real Image Editing

Blended Diffusion for Text-driven Editing of Natural Images

CoLLIE: Continual Learning of Language Grounding from Language-Image Embeddings

LAION-400M: Open Dataset of CLIP-Filtered 400 Million Image-Text Pairs

CLIP-Adapter: Better Vision-Language Models with Feature Adapters

DiffusionCLIP: Text-Guided Diffusion Models for Robust Image Manipulation

Learning to Prompt for Vision-Language Models

ILVR: Conditioning Method for Denoising Diffusion Probabilistic Models

Improving Text-to-Image Synthesis Using Contrastive Learning

Multimodal Few-Shot Learning with Frozen Language Models

Pivotal Tuning for Latent-based Editing of Real Images

Diffusion Models Beat GANs on Image Synthesis

Few-shot Image Generation via Cross-domain Correspondence

StyleCLIP: Text-Driven Manipulation of StyleGAN Imagery

Personalized Federated Learning using Hypernetworks

Learning Transferable Visual Models From Natural Language Supervision

Zero-Shot Text-to-Image Generation

Designing an encoder for StyleGAN image manipulation

GAN Inversion: A Survey

Cross-Modal Contrastive Learning for Text-to-Image Generation

Improved StyleGAN Embedding: Where are the Good Latents?

Denoising Diffusion Implicit Models

DF-GAN: Deep Fusion Generative Adversarial Networks for Text-to-Image Synthesis

Encoding in Style: a StyleGAN Encoder for Image-to-Image Translation

Denoising Diffusion Probabilistic Models

Training Generative Adversarial Networks with Limited Data

Adversarial Latent Autoencoders

In-Domain GAN Inversion for Real Image Editing

Three Approaches for Personalization with Applications to Federated Learning

Personalized Federated Learning: A Meta-Learning Approach

5分で分かる!? 有名論文ナナメ読み：Jacob Devlin et al. : BERT : Pre-training of Deep Bidirectional Transformers for Language Understanding

Image Processing Using Multi-Code GAN Prior

MineGAN: Effective Knowledge Transfer From GANs to Target Domains With Few Images

StarGAN v2: Diverse Image Synthesis for Multiple Domains

Image2StyleGAN++: How to Edit the Embedded Images?

Improving Federated Learning Personalization via Model Agnostic Meta Learning

Semantic photo manipulation with a generative image prior

Image2StyleGAN: How to Embed Images Into the StyleGAN Latent Space?

DM-GAN: Dynamic Memory Generative Adversarial Networks for Text-To-Image Synthesis

Artwork personalization at netflix

Transferring GANs: generating images from limited data

The Unreasonable Effectiveness of Deep Features as a Perceptual Metric

AttnGAN: Fine-Grained Text to Image Generation with Attentional Generative Adversarial Networks

Neural Discrete Representation Learning

GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium

Soft-to-Hard Vector Quantization for End-to-End Learned Compression of Images and Neural Networks

Generative Visual Manipulation on the Natural Image Manifold

Personalized recommender system for e-Learning environment

The conference paper

What's on tv tonight? An efficient and effective personalized recommender system of TV programs

Tooth and Claw

Early Reading Assessment: A Practitioner's Handbook