1. Dialog Normalization (DN) is only for ac3 audio or is also present in other formats like DTS or LPCM?
2. I have noticed with one movie that the music is loud but dialogue is a little quiet. The original is DN -4dB but mine is not set at that level. How can I set to -4 dB with eac3to?
3. ea3to always removes DN, If i can get the original audio file, how can I keep dialog normalization?
+ Reply to Thread
Results 1 to 3 of 3
1. DTS, but I don't know how common it is.
2. I don't think you can.
3. Use the -keepDialnorm option.
DN doesn't have any effect on the dynamic range. Well, I think it can a little in conjunction with any dynamic range metadata, but DN basically just turns everything up or down so the dialogue is at a particular level. That was the idea, to be able to go from one program to the next and the dialogue volume would be the same, but it doesn't have anything to do with how much louder than the dialogue the music might be. I think most software decoders ignore it anyway.
If there's dynamic range compression metadata in the AC3, a Dolby certified decoder should let you adjust it. Some software decoders will. The only two I can think of off the top of my head are ffmpeg and a third party AC3 decoder for foobar2000. The ac3 decoder for foobar2000 only has an option to enable it, which from memory enables the minimum level. ffmpeg lets you adjust it. https://ffmpeg.org/ffmpeg-all.html#AC_002d3-Decoder-Options
A DN of -4dB would actually turn the volume down quite a bit if a decoder was paying attention to it. The target volume for dialogue is -31dB, so if you set -4dB I think you're telling the decoder it needs to decrease the over-all volume by 27dB. It's possible (where do you get the -4dB value?) that -4dB means it's set to the default of -27dB, or there's no DN data. The volume is assumed to be -27dB when no DN range data exists, so without it the volume would be reduced by 4dB to achieve -31dB.
Here's what ffmpeg says:
It uses -31dB as the default, which means no volume change.
Anyway, to apply dynamic range compression permanently, assuming there's metadata for it, you'd have to decode with it enabled and re-encode, and while I only played with it for a couple of files, I wasn't very excited about how the dynamic range compression using the data sounded.
To compress permanently, I ignore all the metadata when decoding and re-encode with foobar2000 while applying compression via a DSP. My favourite is the Dynamic Audio Normalizer, but the command line version doesn't play well with GUIs (long story) however it's also built into ffmpeg, so I use it that way. I have conversion presets saved so it only takes a few seconds to open a file, select a preset, compress and re-encode.
The thread goes on forever, but there's some samples in this post comparing different compressors if you're interested. I can't remember if there's loud music, but it goes from quiet to loud to quiet etc very quickly. I think there's more samples somewhere in the thread.
The 83dB volume mentioned a few times in the post is a volume as it's specified by ReplayGain, a standard for measuring volume. It's meaningless unless you know where it comes from because it's actually a sound pressure level, but it just means the samples were adjusted to the same "average" volume for a fair comparison. In human-speak, they're adjusted to -24dB.
Some more reading:
Last edited by hello_hello; 25th Aug 2019 at 12:45.
Thanks a lot, hello_hello.