A method and apparatus for verifying automatically that a plurality of derivative audio (or other multimedia) files have acceptable sound quality. In one embodiment, each derivative file is compared on a byte-by-byte basis to a corresponding original file to generate a difference. The difference is compared a threshold value (that may be determined empirically). If the difference is too large for many bytes, the derivative file is tagged as having an unacceptable sound quality. In another embodiment, segments of the original and derivative files are converted to the frequency domain and analysis is performed in this domain. The resulting signal could be a tag indicating that whether the derivative file is acceptable, or could be a more comprehensive signal indicative what kind of errors were detected and in what temporal and/or spectral region for diagnostic purposes.