Collecting multimodal image editing requests

A system for creating multimodal image editing requests implemented by a computing device in a digital medium environment, the system comprising:

• a memory to maintain a plurality of image pairs, each image pair including a first image and a second image, the second image having been generated from the first image with an edit of the first image, at least one image pair of the plurality of the image pairs being a placebo image pair used to remove user bias, the placebo image pair having the respective second image that is not generated based on an edit of the respective first image;
• a user interface to sequentially display the plurality of the image pairs by displaying together the respective first image and second image of each image pair;
• a processor system to implement an editing request creation application at least partially in hardware of the computing device, and for each of the image pairs, the editing request creation application is configured to:
• record a multimodal image editing request that includes a user gesture and a voice command that describe the edit of the first image;
• initiate play back of the voice command to generate a user transcription of the voice command;
• generate a data object in a searchable format, the data object including the voice command, the user gesture, and the user transcription;
• and
• train a neural network to recognize the multimodal image editing requests using the data object and the respective first image as inputs of the neural network and the respective second image as an output of the neural network.

In a digital medium environment to create multimodal image editing requests, a method implemented by a computing device, the method comprising:

• displaying a pair of images;
• a step for recording coordinates of a user gesture and audio samples of a voice command, the user gesture and the voice command representing a multimodal image editing request corresponding to the pair of images;
• a step for playing back the audio samples of the voice command from the recording;
• a step for receiving a transcription of the voice command that is user-transcribed from the playing back;
• a step for enabling, responsive to the receiving the transcription, display of a subsequent pair of images;
• a step for generating a data object in a searchable format, the data object including the voice command, the user gesture, and the transcription;
• and
• a step for training a neural network to recognize the multimodal image editing requests using the data object and a first image of the pair of images as inputs of the neural network and a second image of the pair of images as an output of the neural network.