Yitong Wang

Ambiguous object selection in text-guided image editing is solved by directly annotating the canvas, achieving state-of-the-art performance in complex edits.

Direct 2D spatial composition with four binding types unlocks unprecedented controllability in poster generation, outperforming prompt-based methods.

Train a text-to-image model that rivals the state-of-the-art with 1/5th the compute by using GPT-4 to generate better captions.