Feature description
Currently, media generation occurs on first playback of a media file. This takes a long time and is heavily IO bound to the media storage.
Rather than trying to generate this data JIT on first playback, it should be generated when new media is identified. Whether this includes all data (subtitles, attachments, keyframes) or just the super expensive to compute data (keyframes) needs to be discussed further.
Related discord discussion: https://discord.com/channels/1216460898139635753/1216460898609401870/1369583338876178532