Editing images with natural language

UMSI assistant professor Eytan Adar and graduate student Gierad Laput (MSI ’13) collaborated with a team of researchers at Adobe to create a voice-controlled photo editing app called PixelTone.

Read More

Photo editing can be a daunting task with a steep learning curve. Interfaces for photo editing can often be complex, and require the user to learn the language of image processing. Users must understand image properties and learn how they are changed and combined to achieve a desired effect.

Effective image edits are often localized to a specific region and require sophisticated direct manipulations that are typically easier to achieve using the larger displays available on desktop computers.

PixelTone, a multimodal photo editing interface that combines speech and direct manipulation, is designed to make photo editing easier, particularly on the small, portable screens of mobile devices. The application uses natural language to express desired changes to an image, making complex tasks more accessible because the app allows users to simply state a goal without learning complicated interfaces and commands.

Researchers from Adobe were Mira Dontcheva, Gregg Wilensky, Walter Chang, Aseem Agarwala, and Jason Linder.

To read the published results of this project, please click here: http://www.cond.org/pixeltone_compressed.pdf

PixelTone on YouTube


Adobe Research: Editing Images with Natural Language, Adobe Research: $20,000


Adobe Research is committed to developing innovative technologies that will advance Adobe’s mission to change the world through digital experiences.


Press and Web mentions