VisualCloze: A Universal Image Generation Framework via Visual In-Context Learning

[Paper] [Project Page] [Github] [🤗 Dataset Card]

[🤗 Model Card (Diffusers)] [🤗 Model Card (LoRA)]

[💥 New] 👋👋👋 We release the diffusers implementation that makes it easier to use the model through pipelines. See Model Card for details.

📋 License
This project is licensed under apache-2.0.

In-context Example 1

In-context Example 1/1

In-context Example 1/2

In-context Example 1/3

Query

Query 1

Query 2

Query 3

Layout Description (Auto-filled, Read-only)

Task Description (Can be modified by referring to examples to perform custom tasks, but may lead to unstable results)

(Optional) Content Description (Image caption, Editing instructions, etc.)

❗❗❗ Before clicking the generate button, please wait until all images, prompts, and other components are fully loaded, especially when using task examples. Otherwise, the inputs from the previous and current sessions may get mixed.

Number of In-context Examples

0 4

Task Columns

1 5

Generated Results

Task Examples

Each click on a task may result in different examples.

Dense Prediction

Conditional Generation

Image Restoration

Style Transfer

Style Condition Fusion

Virtual Try-On

Relighting

Photodoodle

Editing

Unseen Tasks (May produce unstable effects)

Subject-driven Tasks Examples

Subject-driven Generation

Condition+Subject Fusion

Style Transfer with Subject

Condition+Subject+Style Fusion

Editing with Subject

Image Restoration with Subject

VisualCloze: A Universal Image Generation Framework via Visual In-Context Learning

[💥 New] 👋👋👋 We release the diffusers implementation that makes it easier to use the model through pipelines. See Model Card for details.

In-context Example 1

Query

In-context Example 3

In-context Example 4

In-context Example 5

Task Examples

Subject-driven Tasks Examples