Null-text Inversion for Editing Real Images using Guided Diffusion Models

🤖 Plain-English Summary

Recent large-scale text-guided diffusion models provide powerful image generation capabilities. This allows for keeping both the model weights and the conditional embedding intact and hence enables applying prompt-based editing while avoiding the cumbersome tuning of the model's weights.

🔑 Key Findings

• Currently, a massive effort is given to enable the modification of these images using text only as means to offer intuitive and versatile editing tools.
• To edit a real image using these advanced tools, one must first invert the image with a meaningful text prompt into the pretrained model's domain.
• In this paper, we introduce an accurate inversion technique and thus facilitate an intuitive text-based modification of the image.

🔗 Related Resources