An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion
Rinon Gal1,2, Yuval Alaluf1, Yuval Atzmon2, Or Patashnik1, Amit H. Bermano1, Gal Chechik2, Daniel Cohen-Or1
1Tel Aviv University, 2NVIDIA
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.
This repo contains the official code, data and sample inversions for our Textual Inversion paper.
29/08/2022 Merge embeddings now supports SD embeddings. Added SD pivotal tuning code (WIP), fixed training duration, checkpoint save iterations. 21/08/2022 Code released!
- Release code!
- Optimize gradient storing / checkpointing. Memory requirements, training times reduced by ~55%
- Release data sets
- Release pre-trained embeddings
- Add Stable Diffusion support
Our code builds on, and shares requirements with Latent Diffusion Models (LDM). To set up their environment, please run:
conda env create -f environment.yaml
conda activate ldm
You will also need the official LDM text-to-image checkpoint, available through the LDM project page.
Currently, the model can be downloaded by running:
mkdir -p models/ldm/text2img-large/
wget -O models/ldm/text2img-large/model.ckpt https://ommer-lab.com/files/latent-diffusion/nitro/txt2img-f8-large/model.ckpt
To invert an image set, run:
python main.py --base configs/latent-diffusion/txt2img-1p4B-finetune.yaml
-t
--actual_resume /path/to/pretrained/model.ckpt
-n <run_name>
--gpus 0,
--data_root /path/to/directory/with/images
--init_word <initialization_word>
where the initialization word should be a single-token rough description of the object (e.g., 'toy', 'painting', 'sculpture'). If the input is comprised of more than a single token, you will be prompted to replace it.
Please note that init_word is not the placeholder string that will later represent the concept. It is only used as a beggining point for the optimization scheme.
In the paper, we use 5k training iterations. However, some concepts (particularly styles) can converge much faster.
To run on multiple GPUs, provide a comma-delimited list of GPU indices to the --gpus argument (e.g., --gpus 0,3,7,8)
Embeddings and output images will be saved in the log directory.
See configs/latent-diffusion/txt2img-1p4B-finetune.yaml for more options, such as: changing the placeholder string which denotes the concept (defaults to "*"), changing the maximal number of training iterations, changing how often checkpoints are saved and more.
Important All training set images should be upright. If you are using phone captured images, check the inputs_gs*.jpg files in the output image directory and make sure they are oriented correctly. Many phones capture images with a 90 degree rotation and denote this in the image metadata. Windows parses these correctly, but PIL does not. Hence you will need to correct them manually (e.g. by pasting them into paint and re-saving) or wait until we add metadata parsing.
To generate new images of the learned concept, run:
python scripts/txt2img.py --ddim_eta 0.0
--n_samples 8
--n_iter 2
--scale 10.0
--ddim_steps 50
--embedding_path /path/to/logs/trained_model/checkpoints/embeddings_gs-5049.pt
--ckpt_path /path/to/pretrained/model.ckpt
--prompt "a photo of *"
where * is the placeholder string used during inversion.
LDM embedding checkpoints can be merged into a single file by running:
python merge_embeddings.py
--manager_ckpts /path/to/first/embedding.pt /path/to/second/embedding.pt [...]
--output_path /path/to/output/embedding.pt
For SD embeddings, simply add the flag: -sd or --stable_diffusion.
If the checkpoints contain conflicting placeholder strings, you will be prompted to select new placeholders. The merged checkpoint can later be used to prompt multiple concepts at once ("A photo of * in the style of @").
Datasets which appear in the paper are being uploaded here. Some sets are unavailable due to image ownership. We will upload more as we recieve permissions to do so.
Pretained models coming soon.
Stable Diffusion support is a work in progress and will be completed soon™.
- Adding "a photo of" to the prompt usually results in better target consistency.
- Results can be seed sensititve. If you're unsatisfied with the model, try re-inverting with a new seed (by adding
--seed <#>to the prompt).
If you make use of our work, please cite our paper:
@misc{gal2022textual,
doi = {10.48550/ARXIV.2208.01618},
url = {https://arxiv.org/abs/2208.01618},
author = {Gal, Rinon and Alaluf, Yuval and Atzmon, Yuval and Patashnik, Or and Bermano, Amit H. and Chechik, Gal and Cohen-Or, Daniel},
title = {An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion},
publisher = {arXiv},
year = {2022},
primaryClass={cs.CV}
}
Here are some sample results. Please visit our project page or read our paper for more!
