Datasets:

Modalities:
Image
Text
Languages:
English
ArXiv:
Libraries:
Datasets
License:
Dataset Viewer
Auto-converted to Parquet
image
imagewidth (px)
512
9k
instruction
stringlengths
3
263
category
stringclasses
5 values
key
stringlengths
7
7
image_idx
stringclasses
108 values
prompt_idx
stringclasses
24 values
make the cat very fat
Instruction Editing - Local
0000_00
0000
00
give the cat a tophat
Instruction Editing - Local
0000_01
0000
01
the camera zoomed out and the cat is now dancing on its hind legs
Character Reference
0000_02
0000
02
turn the cat into a dog
Instruction Editing - Local
0000_03
0000
03
a photo of this cat swimming in a river in a jungle, go pro action footage
Character Reference
0000_04
0000
04
an old victorian era painting of this cat on a throne
Character Reference
0000_05
0000
05
make the cat orange
Instruction Editing - Local
0000_06
0000
06
replace the carpet with a wood floor
Instruction Editing - Local
0000_07
0000
07
replace the environment with a field of flowers, the cat is laying in the middle of the flowers
Character Reference
0000_08
0000
08
Edit the word "Lover" to "Hater"
Text Editing
0001_00
0001
00
Add a second thumb to the hand
Instruction Editing - Local
0001_01
0001
01
Turn the hand upside down, keep the text as is
Instruction Editing - Local
0001_02
0001
02
Turn this into a neon sign hanging on a brick wall in a cool modern office
Instruction Editing - Global
0001_03
0001
03
Place this on a glass door in front of a building
Character Reference
0001_04
0001
04
Remove the text
Instruction Editing - Local
0001_05
0001
05
Remove the hand in the middle
Instruction Editing - Local
0001_06
0001
06
Change the color to purple
Instruction Editing - Global
0001_07
0001
07
Change the background to purple
Instruction Editing - Global
0001_08
0001
08
Change the text to green and the hand to purple
Instruction Editing - Local
0001_09
0001
09
An image of a street with a dozen posters of this all over a fence
Character Reference
0001_10
0001
10
make all the signs green
Instruction Editing - Local
0002_00
0002
00
remove the two green signs
Instruction Editing - Local
0002_01
0002
01
make the black sign text say "Pick a new direction"
Text Editing
0002_02
0002
02
change the text "Singapore" to "Freiburg"
Text Editing
0002_03
0002
03
Remove all the signs and unblur the background
Instruction Editing - Global
0002_04
0002
04
Make "HongKong" say "King Kong"
Text Editing
0002_05
0002
05
Add a monkey on the green sign above the word "Kuala"
Instruction Editing - Local
0002_06
0002
06
Every sign now says "Why not this way"
Text Editing
0002_07
0002
07
Remove the man in the brown jacket
Instruction Editing - Local
0003_00
0003
00
Remove the man in the blue jacket
Instruction Editing - Local
0003_01
0003
01
This image in the style of Van Gogh's "The Starry Night"
Instruction Editing - Global
0003_02
0003
02
Remove the sun but keep the brightness in the image
Instruction Editing - Local
0003_03
0003
03
Replace the snow with a forest
Instruction Editing - Global
0003_04
0003
04
There is now a volcano on the mountain behind them erupting
Instruction Editing - Local
0003_05
0003
05
Turn both of the men into storm troopers
Instruction Editing - Local
0003_06
0003
06
Zoom in on the mans face closest to the camera
Instruction Editing - Global
0003_07
0003
07
Make the first mans jacket red
Instruction Editing - Local
0003_08
0003
08
Add a horse of ogres running into battle on the mountain behind them
Instruction Editing - Local
0003_09
0003
09
The camera shifted to in front of them now showing both of them on the chair lift from the front
Character Reference
0003_10
0003
10
Zoom out to show the whole scene
Instruction Editing - Global
0003_11
0003
11
Make this into a renaissance painting
Instruction Editing - Global
0003_12
0003
12
Turn this into an expressionist art piece
Instruction Editing - Global
0003_13
0003
13
Add a small weird creature between them
Instruction Editing - Local
0003_14
0003
14
Mirror the image so its symmetrical down the middle
Instruction Editing - Global
0003_15
0003
15
Make it darker
Instruction Editing - Global
0003_16
0003
16
Remove the brightness
Instruction Editing - Global
0003_17
0003
17
Make a new image, its just the color green and some triangles
Instruction Editing - Global
0003_18
0003
18
Its now nighttime, replace the sun with the moon
Instruction Editing - Global
0003_19
0003
19
The man raises his hands into the sky
Instruction Editing - Local
0004_00
0004
00
The same man is now sitting on a throne
Character Reference
0004_01
0004
01
Remove the house in the background
Instruction Editing - Local
0004_02
0004
02
Remove the skis
Instruction Editing - Local
0004_03
0004
03
The man is now standing on the skis and riding away
Instruction Editing - Local
0004_04
0004
04
Add a cheerful snowman wearing a striped scarf sitting next to him
Instruction Editing - Local
0004_05
0004
05
Make this image into a crayon drawing
Instruction Editing - Global
0004_06
0004
06
Make it bright daytime
Instruction Editing - Global
0004_07
0004
07
Make it night time
Instruction Editing - Global
0004_08
0004
08
Its now spring season with flowers blooming on the sidewalk
Instruction Editing - Global
0004_09
0004
09
Make it summer
Instruction Editing - Global
0004_10
0004
10
Remove all the snow
Instruction Editing - Global
0004_11
0004
11
Make a 4 panel image showing this location in winter (current), spring, summer, and fall
Instruction Editing - Global
0004_12
0004
12
Add a cute drawing of a deer into the snow
Instruction Editing - Local
0004_13
0004
13
Fill the image with deer
Instruction Editing - Global
0004_14
0004
14
Add a bus driving on the road
Instruction Editing - Local
0004_15
0004
15
Add a small silver robot companion sitting next to him
Instruction Editing - Local
0004_16
0004
16
Change this to a cowboy hat
Instruction Editing - Local
0005_00
0005
00
Remove this
Instruction Editing - Local
0005_01
0005
01
Remove the object in the red square
Instruction Editing - Local
0005_02
0005
02
Remove the red square only
Instruction Editing - Local
0005_03
0005
03
This duck on a bike in a forest
Character Reference
0005_04
0005
04
Make this into anime artwork
Instruction Editing - Global
0005_05
0005
05
Using this style make some art of a boat in the harbor of an old town
Style Reference
0006_00
0006
00
Create art of a brutalist concrete building in a field under the moon using this exact art style
Style Reference
0006_01
0006
01
remove the bed
Instruction Editing - Local
0006_02
0006
02
add a dog sitting on the chair on the left
Instruction Editing - Local
0006_03
0006
03
A photo of my messy bedroom. this painting is on the wall
Instruction Editing - Global
0006_04
0006
04
turn this into a real photo
Instruction Editing - Global
0006_05
0006
05
The camera moves forward and looks right showing more of the bed and the paintings on the wall
Instruction Editing - Global
0006_06
0006
06
open the window showing a field outside
Instruction Editing - Local
0006_07
0006
07
make the walls red
Instruction Editing - Local
0006_08
0006
08
turn her into a bear
Instruction Editing - Global
0007_00
0007
00
make her old
Instruction Editing - Local
0007_01
0007
01
she is now riding a bike through a forest on a dirt path
Character Reference
0007_02
0007
02
this girl sailing near a lighthouse
Character Reference
0007_03
0007
03
she is now studying in a dimly lit room at night
Character Reference
0007_04
0007
04
zoom out showing her standing
Character Reference
0007_05
0007
05
make her dance
Character Reference
0007_06
0007
06
using this style create a duck in a tuxedo
Style Reference
0007_07
0007
07
change her eyes to red
Instruction Editing - Local
0007_08
0007
08
make her bald
Instruction Editing - Local
0007_09
0007
09
make her do heart hands
Instruction Editing - Local
0007_10
0007
10
Make her into a real person
Character Reference
0007_11
0007
11
turn the object to face the camera directly
Instruction Editing - Local
0008_00
0008
00
place this object on a marble pedestal in a museum inside of a glass display case
Character Reference
0008_01
0008
01
put this object on the dashboard of a dirty beat up car, interior pov photo
Character Reference
0008_02
0008
02
place the logo on the object onto a tshirt being worn by a muscular man
Character Reference
0008_03
0008
03
a table with many of these objects on it
Instruction Editing - Global
0008_04
0008
04
change the text on the object to say "Cool Little Easel"
Text Editing
0008_05
0008
05
put a picture of a clown on the object
Instruction Editing - Local
0008_06
0008
06
Change the text from "Black Forest Labs" to "Make Pixels Beautiful"
Text Editing
0009_00
0009
00
End of preview. Expand in Data Studio

Kontext Bench

Kontext Bench is a benchmark for image editing models consisting of source images paired with image editing instructions and category tags.

The benchmark comprises 1026 unique image-prompt pairs derived from 108 base images from diverse sources. It spans five core tasks: local instruction editing, global instruction editing, text editing, style reference, and character reference. We found that the scale of the benchmark provides a good balance between reliable human evaluation and comprehensive coverage of real-world applications.

Benchmark Structure

kontext-bench/
└── test/
    ├── images/
    │   ├── 0000.jpg
    │   ├── 0001.jpg
    │   └── ...
    └── metadata.jsonl

Fields

Each line in metadata.jsonl consists of

  • file_name: Path to the image file relative to the split directory
  • instruction: The editing instruction to apply to the image
  • category: Category of the editing instruction
  • key: Unique identifier for this image-instruction pair
  • image_idx: Index of the source image
  • prompt_idx: Index of the prompt for this image

An example entry is shown below:

{
    "file_name": "images/0000.jpg"
    "instruction": "give the cat a tophat",
    "category": "Instruction Editing - Local",
    "key": "0000_01",
    "img_idx": "0000",
    "prompt_idx": "01"
}

Benchmark Statistics

  • Total entries: 1026
  • Unique images: 108

Category Statistics

  • Character Reference: 193 entries
  • Instruction Editing - Global: 262 entries
  • Instruction Editing - Local: 416 entries
  • Style Reference: 63 entries
  • Text Editing: 92 entries

License

The benchmark is released under the MIT License. This benchmark and the included Images are made available for scientific and research purposes only. We gratefully acknowledge all contributing photographers, Unsplash, Pexels for making their visuals available to the research community.

Citation

@misc{labs2025flux1kontextflowmatching,
      title={FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space}, Add commentMore actions
      author={Black Forest Labs and Stephen Batifol and Andreas Blattmann and Frederic Boesel and Saksham Consul and Cyril Diagne and Tim Dockhorn and Jack English and Zion English and Patrick Esser and Sumith Kulal and Kyle Lacey and Yam Levi and Cheng Li and Dominik Lorenz and Jonas Müller and Dustin Podell and Robin Rombach and Harry Saini and Axel Sauer and Luke Smith},
      year={2025},
      eprint={2506.15742},
      archivePrefix={arXiv},
      primaryClass={cs.GR},
      url={https://arxiv.org/abs/2506.15742},
}
Downloads last month
0