title
Products            Buy            Support Forum            Professional            About            Codec Central
 

AAC/M4A VBR Q1 producing larger files than Q2

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • Citizen32811
    • Nov 2022
    • 2

    AAC/M4A VBR Q1 producing larger files than Q2

    Compressing a selection of 9 language lesson files (total 176MB FLAC) with no filters results in the following total size and bitrate range (low-median-high).

    Q1 28.5MB 46-52-83kbps
    Q2 27.9MB 46-51-80kbps
    Q3 34.6MB 57-64-96kbps
    Q4 42.9MB 71-79-134kbps
    Q5 65.9MB 111-122-207kbps

    I want a conversion product I can trust; this doesn't seem like the best start. It seems unlikely to expect a convincing answer that doesn't involve a bug, but you can try.

    -Mr. Audio
  • Spoon
    Administrator
    • Apr 2002
    • 44095

    #2
    Re: AAC/M4A VBR Q1 producing larger files than Q2

    You miss understand how lossy compression works, based on frequency dropping (psychoacoustic analysis), which will be tuned to music rather than voice. It does not surprise me the results you get, and it does not mean there is a bug.
    Spoon
    www.dbpoweramp.com

    Comment

    • Lynx_TWO
      dBpoweramp Enthusiast
      • Aug 2009
      • 83

      #3
      Re: AAC/M4A VBR Q1 producing larger files than Q2

      What Spoon said - lossy compression involves a lot of maths and psychoacoustic analysis. If there IS a bug it's in the Fraunhofer FDK AAC encoder itself, not in the dBpoweramp software since the software is basically a (very nice) frontend for other external codes. I would report these results over here: https://github.com/mstorsjo/fdk-aac/issues as this is the actual repository for that particular codec.

      I've been using dBpoweramp for over a decade and I can tell you it is extremely reliable and robust software.

      If you are looking to host podcast files, I would suggest a codec that is tuned for voice AND music such as OPUS (also royalty-free unlike AAC)
      Last edited by Spoon; December 02, 2022, 10:01 AM.

      Comment

      • Citizen32811
        • Nov 2022
        • 2

        #4
        Re: AAC/M4A VBR Q1 producing larger files than Q2

        A reply starting with a statement that could be easily interpreted as a put-down may be a less than ideal way to begin. However, your observations regarding my technical knowledge of psychoacoustic analysis are correct. I would offer that some people have people skills to the exclusion of technical skills while others have the reverse. As you are clearly a technical person here to help others, may I this offer this sample reply for future consideration. "We understand how your results don't seem intuitive to a casual user and could cause concern." This expression of empathy and understanding is customer service 101.

        In response, obviously providing more-lossy compression is a tradeoff that users are accepting with the expectation of saving more space. While the unexpected results may not be an indication of a bug, you may agree that if they are consistent with a wide variety of samples, they could be a bug. In any case a larger file with a lower quality score is clearly not what the customer is looking for. Meeting a customer's expectations and convincing potential customers that your product or service will meet their expectations is a key to the successful marketing of any product or service.

        Kindly consider seriously that this unexpected behavior, although "not indicative of a bug" it is still likely to dissatisfy users who have opted for more loss for a smaller file and have instead received a larger file.

        Thank you for your kind attention.
        Last edited by Citizen32811; December 04, 2022, 06:53 PM.

        Comment

        • GBrown
          dBpoweramp Enthusiast
          • Oct 2009
          • 290

          #5
          Re: AAC/M4A VBR Q1 producing larger files than Q2

          Originally posted by Citizen32811
          I want a conversion product I can trust; this doesn't seem like the best start. It seems unlikely to expect a convincing answer that doesn't involve a bug, but you can try.
          I think you may have set the bar here for a response such as what was provided. Nonetheless the reply holds merit with regards to your concern. Lossy compression for audio files are built around music, which tends to have wider bandwidths to deal with. Speech files have a very limited range, and as such it is reasonable to get results like this once you are looking at the higher compression ranges that have different algorithms to focus on. A more reasonable test to confirm this would be to use a file with a more complex arrangement of music, where you are more likely to see the expected results. A few hundred thousand or so other users would have otherwise complained about this anomaly long ago.

          Comment

          • Spoon
            Administrator
            • Apr 2002
            • 44095

            #6
            Re: AAC/M4A VBR Q1 producing larger files than Q2

            Probably millions using the fdk acc encoder world wide.
            Spoon
            www.dbpoweramp.com

            Comment

            • garym
              dBpoweramp Guru
              • Nov 2007
              • 5773

              #7
              Re: AAC/M4A VBR Q1 producing larger files than Q2

              ...and keep in mind that @Lynx_TWO and @GBrown are both just regular users of the program (like me) trying to assist with posted questions, as many of us on this forum try to do. We are not customer service employees. @Spoon is the developer/owner of the product.

              Comment

              • Lynx_TWO
                dBpoweramp Enthusiast
                • Aug 2009
                • 83

                #8
                Re: AAC/M4A VBR Q1 producing larger files than Q2

                Originally posted by GBrown
                A few hundred thousand or so other users would have otherwise complained about this anomaly long ago.
                Well, in all fairness to the OP, I kinda doubt that most users (G-d I hate that word, we seem to only use “users” for drug addicts and people who use software lol) actually test things like he (she?) did… but yes, the way the OP presented seemed a bit confrontational… but then again it’s really easy to take things in the worst possible way (kinda like how my recent ex did :/ ) over text… However, it’s totally possible that there is some kind of bug in the original encoder… When I get home, I’ll do some tests on my end with music and see if I can replicate the issue.

                Comment

                Working...

                ]]>