Figure 1. The Legion Scribe system. Audio is captured from a user's mobile device or laptop and streamed to a group of non-expert workers, who each caption what they can from the audio. Then these partial captions are combined into a single caption and sent back to the user.