MagicBrush: A Manually Annotated Dataset for Instruction-Guided Image Editing

By Kai Zhang, Lingbo Mo, Wenhu Chen, Huan Sun, Yu Su. Neural Information Processing Systems, 2023

689 citations recorded by the source corpus.

Text-guided image editing is widely needed in daily life, ranging from personal use to professional applications such as Photoshop. However, existing methods are either zero-shot or trained on an automatically synthesized dataset, which contains a high volume of noise. Thus, they still require lots of manual tuning to produce desirable outcomes in practice. To address this issue, we introduce MagicBrush (https://osu-nlp-group.github.io/MagicBrush/), the first large-scale, manually annotated dataset for instruction-guided real image editing that covers diverse scenarios: single-turn, multi-turn, mask-provided, and mask-free editing. MagicBrush comprises over 10K manually annotated triplets (source image, instruction, target image), which supports trainining large-scale text-guided image editing models. We fine-tune InstructPix2Pix on MagicBrush and show that the new model can produce much better images according to human evaluation. We further conduct extensive experiments to evaluate current image editing baselines from multiple dimensions including quantitative, qualitative, and human evaluations. The results reveal the challenging nature of our dataset and the gap between current baselines and real-world editing needs.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/ecf38d702bea560c1f9d372645e2be7661a71f37

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space (2025, positive): “MagicBrush [60], while using authentic MS-COCO images, is constrained by DALLE-2’s [41] capabilities during data collection.”
Step1X-Edit: A Practical Framework for General Image Editing (2025, neutral): “Early approaches such as InstructEdit [55], InstructPix2Pix [8], MagicBrush [66], and BrushEdit [28] adopt modular pipelines where MLLMs generate prompts, spatial cues, or synthetic instruction-image pairs to guide diffusion-based editing.”
VACE: All-in-One Video Creation and Editing (2025, neutral): “…[9, 22, 64] generation in the field, there has been a proliferation of down-stream tasks and applications, such as repainting [3, 82], editing [4, 42, 68, 70, 75], controllable generation [30, 76], frame reference generation [20, 73], and ID-referenced video synthesis [11, 35, 47, 74].”
Emu Edit: Precise Image Editing via Recognition and Generation Tasks (2023, positive): “Instruction-based image editing [2, 25] attempts to resolve these limitations by allowing users to effortlessly describe their editing goals using natural language instructions.”
Qwen-Image Technical Report (2025, neutral): “Janus-Pro (Chen et al., 2025b) 0.019 0.006 BLIP3-o (Chen et al., 2025a) 0.021 0.018 Kolors 2.0 (team, 2025) 0.258 0.329 BAGEL (Deng et al., 2025) 0.373 0.310 OmniGen2 (Wu et al., 2025b) 0.561 0.059 X-Omni (Geng et al., 2025) 0.900 0.814 HiDream-I1-Full (Cai et al., 2025) 0.543 0.024 FLUX.1 [Dev] (Bl”
OmniGen: Unified Image Generation (2024, neutral): “We collect the data from multiple tasks and sources: image editing (MagicBrush [74], InstructPix2Pix [3] and SEED-edit [17], ), human motion (Something-Something [21]), virtual try-on (HR-VITON [30] and FashionTryon [77]), and style transfer (StyleBooth [23]).”
SEED-X: Multimodal Models with Unified Multi-granularity Comprehension and Generation (2024, neutral): “We fine-tune the visual de-tokenizer on MagicBrush [30] and in-house image editing data, as well as the pure images in the first stage, where the conditional inputs are set to zeros.”