moodboard: macOS viewer for AI images, video, audio

Moodboard began as a viewer written to compare generated images while building an X-participation-based battle game. It was then split out as an independent tool, with video, audio, and selection modes added. That origin explains the file types it handles and the selection flow: when a single agent request returns several image or audio candidates, users need to look at them together rather than open files one by one.

The tool reuses the same window for the same task, with new results replacing old ones, and gives windows names so separately running projects or agents stay apart. The agent only reports "it opened" after confirming the window has actually finished loading files, and it also surfaces files that failed to open, so a failed display is not reported as success.

On the window itself, users enlarge cards under the cursor with the wheel or pinch, drag them, click for full size (1 for 1x, 2 for 200%, arrow keys to move), toggle the background with b, and restore the original arrangement with r. The interface and error messages follow the macOS language setting in Japanese and English. The main limits are that only macOS has been verified, and .webm with transparent VP9 may not play, though .mov with transparent HEVC does.