No announcement yet.

Audio Normalization Issues

  • Filter
  • Time
  • Show
Clear All
new posts

  • Audio Normalization Issues

    I have an hour long audio track in my project. I am splitting it up at specific points and then trying to normalize each individual section. I noticed one section, which was rather quiet, was only normalizing with a gain of 1.1. I set the gain of that section back to 0, exported just that section, and then imported the resulting wave file back in. The gain of the section is now registering as 17.

    To sum up:

    Split clip from bigger audio track normalizes with a gain of 1.1
    Export clip and import new wave. Now normalizes with a gain of 17.

  • #2
    the normalize tool act strange sometimes because it is peak based rather than RMS based like Soundforge

    so all you need is a 1 frame spike and normalize will give you a strange result if performing after the clip is cut

    so what I do:


    1. I view linear waveforms and not log
    2. I increase audio track height
    3. cut up the clip
    4. normalize the first until the mixer shows me the level I am after, this may need several value changes until I hit the mark
    5. I remember this value, then normalize the next clip with the same value
    6. if the waveform does not look the same height as my reference clip, I will normalize again with a different value until the waveform is the same height


    export entire wave as new file, open in Soundforge and normalize using RMS option to -10db, then apply SF limiter to -6db (no compression)

    the limiter will dynamically compress spikes to -6db

    then save as new file and add back to EDIUS and cut it up any way you wish
    Anton Strauss
    Antons Video Productions - Sydney

    EDIUS X WG with BM Mini Monitor 4k and BM Mini Recorder, Gigabyte X299 UD4 Pro, Intel Core i9 9960X 16 Core, 32 Threads @ 4.3Ghz, Corsair Water Cooling, Gigabyte RTX-2070 Super 3X 8GB Video Card, Samsung 860 Pro 512GB SSD for System, 8TB Samsung Raid0 SSD for Video, 2 Pioneer BDR-209 Blu-ray/DVD burners, Hotswap Bay for 3.5" Sata and 2.5" SSD, Phanteks Enthoo Pro XL Tower, Corsair 32GB DDR4 Ram, Win10 Pro 21H2


    • #3
      This may be of interest, from a similar post.

      Be careful if applying any dynamic effects globally across different clips that are from different takes that may have different shoot settings. Depending on source/shoot setup applying dynamics across multiple clips wont give an even outcome across all of them.


      • #4
        Its too bad they didn't have a peak based option. I get what antonsvideo is saying about audio but that still doesn't explain why the exact same audio clip behaves differently when exported and imported on its own, vs being a clip from a bigger audio file.


        • #5
          It's possible the clip is being peak-normalized based on the entire clip, rather than the trimmed clip.


          • #6
            That was what I suspected as well, however other clips from the same audio file are normalizing to different values. Part of me almost wonders if there's a bug somewhere in this.


            • #7
              Odd. Any node/volume adjustments on that clip?


              • #8
                No. What I have is a hour long clip of a performance. Some of the dances in the performance are quieter then others so I am splitting the clip or cutting it at silent points in between dances and then trying to normalize each section as a starting point. The first 2 sections went fine but then this quieter 3rd section wasn't adjusting much at all which is what alerted me to the problem.


                • #9
                  <-- is puzzled. :\


                  • #10
                    I can potentially try to reproduce the problem on a smaller chunk of the audio file and then share it off. I'll see if I cannot do that tonight. At the moment this feels like a bug of some kind.


                    • #11
                      I have had good success with exporting the whole timeline as a WAV file then using Levelator (free download) on that clip and re-importing that file as my audio back into my project. The whole timeline now had a nice uniform audio level.
                      i7 3930K, Win 10 pro 64 Bit.
                      RAM 32 GB.
                      ASUS P9X79-Deluxe.
                      PS Thermaltake 1kW.
                      Graphics Gigabyte GTX1660 OC.
                      OS:Intel 535 SSD 480 GB, Data: two 2TB WD Cavier Black, RAID 0.
                      Cameras: Sony EX1, AX100, x1000v, A7R4, DJI Mavic 2 Pro
                      SW: Edius v8.53 WG & 9.52 WG, Vegas 17 Pro, CS6, Resolve Studio 16, Hitfilm 2017, Sound Forge 13, SpectraLayers Pro v6.0, Neat Video v5, Vistitle 2.8.5 TMPEnc VMWorks 7, Auth. Works 6, MochaPro v4.1, Robuskey, DVD Architect 7, 3D LUT Creator.


                      • #12
                        I am not interested in creating copies of my audio tracks (they are large files) and importing them back in. I'd prefer to be able to do as much as possible from within the editor itself. Its nice to know about Levelator but i prefer a level of manual control, especially since this is a dance performance involving music.


                        • #13
                          Just curious, does the behavior change if you embed that clip (as a sequence) into another sequence?


                          • #14
                            I may be alone in this but I regard "normalising" filters and similar as some of the most useless software gizmos around.

                            In my view not only is sound mixing at least as important as video editing but it should be done second by second by hand and ear with one eye on the meters. Plonking a filter on a clip and hoping for the best gives you what you deserve. Why not just shove the rubber band up or down?

                            The fact that Edius users are wasting time applying these frankly amateur filters when there's a built in interface in Edius for an inexpensive control surface (Behringer BCF2000) makes it even more surprising.

                            I don't mean to hurt people's sensitivities but when we have professional tools at our fingertips in Edius why complain when we don't use them?
                            Edius 6.52; i2500K; Gigabyte Z68P-DS3; 8Gb RAM; 124Gb SSD for system, 5x1Tb HDD on board; 1Tb & 500Gb HDD off-board; 2Tb RAID1 out-board; BCF2000; Shuttle Pro2; Radeon 5 series dual display, HD Spark, HD monitor. Windows 7 Pro 64-bit, New Blue Fx Art Blends, Video Essentials 2, 3, 4, Titler Pro 1 and ColourFast, ProDad Defishr.


                            • #15
                              @Philip Howells

                              Perhaps there is some mis-understanding about what the normalization filter does. The normalization filter I am most familiar with will automatically raise the sound level of the audio level such that the maximum sound made is at the level you ask for. As long as you choose the maximum level wisely, you can easily raise the volume without clipping, or changing the sound in any way.

                              This is what I want as my starting point. I want to raise the volume of my sound clip, without modifying it in any other way, such that the loudest sound in the clip is at a given level. From there I will proceed to modify it as I see fit.

                              Unfortunately, the normalizing filter in EDIUS doesn't look at the peak level, it uses some RMS stuff and it apparently doesn't seem to analyze smaller clips properly.

                              I never want to clip my sound track and if I were to adjust the volume level manually this could definitely happen.

                              Its nice to know there exists a device (Behringer BCF2000) out there and it doesn't look that expensive. It is something I'll have to think about in the future.