Few people seem to be interested in eliminating dupes by comparing the 'audio content' less all tags. AFAIK, there is no third party app yet that will scan and compare only the content, disregarding all tags, because of the way tags are attached to either the beginning or end, or both, of the audio.
Since I have about 160,000 tunes, about 20,000 of which are (were) dupes or trips or more, I have a vital interest in dumping exact dupes, and use a multi-step process as a successful process for most, but not all, duplicates.
1. I write something to a tag (I use a # in the access rating tag)--this insures that there are id3v2.3 tags
2. I delete the id3v1 tags
3. I clean and re-write tunes to name-artist-album filenames, using symbols (+++) to signify to the dupe finder that the files have no v1 tags
4. I scan with DoubleKiller, having arrived at the settings that work most effectively for me:
In Comparison Options, I check size with a tolerance of plus/minus 4kb, and check content kbytes 31 to 1540, seen from end of file, and fingerprint (crc32).
Theoretically the audio has no tags at the end anymore...I have played around for many hours to get this far, but there are still 'similar' or apparently 'identical' files I can't seem to filter out, but are apparent in MC. The .chm help file with DoubleKiller discusses techniques in detail, but actually isolating the audio for scanning is apparently rocket science for programmers...
Since I convert all .flac and other formats to .mp3 before adding to my processing and my library, this may contribute to the difficulties. I have never seen comments regarding how bitrate and conversion affects crc.
Even though the 'duration' values in MC are in seconds there are unexplainable wide inconsistencies that lead me to believe that MC duration is seldom a reliable guide for dupe finding...maybe .mp3 bitrates have a significant effect on the duration. I have tried re-converting .mp3s to a uniform bitrate but there is no change in MC duration.