Increasing levels in an AC3 track

[carlmart]

Which is the best and cleanest way to increase levels in an AC3 track?

This time I would like to preserve the original mix. Usually on such cases I convert the AC3 to wav and adjust levels in SoundForge. Then convert back to AC3. But I lose the 5.1 mix.

From what I've seen it would need increasing levels by 8.3 dB. What should I use to do that within AC3?

[poisondeathray]

I'm pretty sure you can export 5.1 AC3 in soundforge

You can even do it in audacity (you just have to set it up in the preferences)

[carlmart]

SoundForge 8.0 does not read or import or can't do anything with more than two tracks.

[poisondeathray]

SoundForge 8.0 does not read or import or can't do anything with more than two tracks.

That's weird. Even the free audacity can handle that. Are you sure it's not setting or plugin that you are missing?

Other free options (if you know the exact gain) to apply are eac3to, foobar2k, avisynth+aften or soundout ... etc...

[carlmart]

OK. I am using Soundout with Avisynth to split the AC3 tracks.

How do I adjust levels with it?

[Soopafresh]

https://forum.videohelp.com/topic317891.html

[poisondeathray]

OK. I am using Soundout with Avisynth to split the AC3 tracks.

How do I adjust levels with it?

Why are you splitting them? Do you mean GetChannels() ? Do you need to adjust them separately?

AmplifyDB() is what you want

^Soopa's method looks easy as well

[carlmart]

OK. I am using Soundout with Avisynth to split the AC3 tracks.

How do I adjust levels with it?

Why are you splitting them? Do you mean GetChannels() ? Do you need to adjust them separately?

AmplifyDB() is what you want

^Soopa's method looks easy as well

I am splitting the film so I can get higher bitrate levels in my conversion.

Sorry. Which is Soopa's method?

[poisondeathray]

He posted a link for a batch file method (look 3 posts up) , just another option

[carlmart]

I am trying Soopa's method as we speak.

But shouldn't it be better to do that in one single conversion with soundout?

[poisondeathray]

I am trying Soopa's method as we speak.

But shouldn't it be better to do that in one single conversion with soundout?

I haven't used Soopa's method, but doesn't that work as a single conversion as well?

If the settings used were the same, with the same aften build, then I think you would get the same results

[Soopafresh]

My method isn't necessarily better, but it is fast and brainless.

Rather than using soundout, I use a command line utility to transcode the AVS script and send the output to the very HQ Aften AC3 encoder.

[raquete]

Which is the best and cleanest way to increase levels in an AC3 track?

This time I would like to preserve the original mix.

From what I've seen it would need increasing levels by 8.3 dB. What should I use to do that within AC3?

best and cleanest that will preserve the original mix is:
encrease 8.3 dB(level/volume) in your amplifier/receiver.

all the remainders ways works but are lossy over lossy...i mean,
you will decode one lossy source, encrease volume and encode in lossy format again...
result? lossy more lossy with high level and you don't want it, right?

[carlmart]

[quote="raquete"]

best and cleanest that will preserve the original mix is:
encrease 8.3 dB(level/volume) in your amplifier/receiver.

all the remainders ways works but are lossy over lossy...i mean,
you will decode one lossy source, encrease volume and encode in lossy format again...
result? lossy more lossy with high level and you don't want it, right?

The level I mentioned would allow me to keep the original audo dynamics, because dialog is much much lower. Something like 20 or 30dB lower. So if you add up it's 28 or 38dB lower, which I think it's a lot to compensate with amp setting.

On some (rare) films the higher level peaks are not so much higher than the dialogue levels. In those films I can apply a higher level setting and things will not be so different from "usual" levels.

In this case, as I am splitting the audio track, I wondered how I could use the same operation to set higher levels. My concern was also about re-processing an already processed track. Can't I do the splitting and volume setting on only one pass with soundout?

I also would like some more predictable (by me) volume correction than Soopa's method (very practical indeed) allows.

[Soopafresh]

Have a look at this for a method to change the volume of a single channel

http://avisynth.org/GetChannel

Have a look at this thread for specific examples of the command

https://forum.videohelp.com/topic326127.html

[poisondeathray]

Soopa: I think he means splitting as in cutting into sections like A+B

carlmart - why not use Normalize() ? This will give you predictable results without clipping

If you want to split your video, you can use Trim(). I think you need 2 scripts if you want 2 sections. Another method is to encode it as 1 and use delaycut after.

If you want to normalize each channel, use GetChannel() with it

If you need finer control over sections, you can use audacity, and export to 5.1 AC3

[Alex_ander]

SoundForge 8.0 does not read or import or can't do anything with more than two tracks.

SoundForge 9 opens DD5.1 as a part of VOB (in multi-track mode), then video can be detached. This only does not work if the ac3 stream has wrong header. In case of DD2.0 it even works for ac3's just renamed to .mpg or .VOB.

[carlmart]

Soopa: I think he means splitting as in cutting into sections like A+B

carlmart - why not use Normalize() ? This will give you predictable results without clipping

Yes, I split the film in sections or parts, usually to improve my bitrates. When the film is important to me, I try to see above 6000 kbit/s on my conversion to DVD. It may sound unlikely, but the result shows up on my 720p plasma screen.

I am learning on this normalize setting. It seems very interesting:

http://avisynth.org/oldwiki/index.php?page=Normalize

If you want to split your video, you can use Trim(). I think you need 2 scripts if you want 2 sections. Another method is to encode it as 1 and use delaycut after.

Yes, I already use trim(), both for video conversion and audio. My next amp will probably have DTS, so I won't need to convert to AC3 anymore.

If you want to normalize each channel, use GetChannel() with it

If you need finer control over sections, you can use audacity, and export to 5.1 AC3

I will try audacity. Apparently you can import 5.1 Soundforge 9, but with vob. In my case I have mkv files with AC3 or DTS audio tracks, and they need level adjust very often.

[carlmart]

why not use Normalize() ? This will give you predictable results without clipping

Don't seem to quite get how to write this script.

I have an AC3 file that I want to normalize and still get an AC3 file output.

On the normalize() literature I found what might be an example on what I want to do.

video = AviSource("C:\video.avi")
audio = WavSource("c:\autechre.wav")
audio = Normalize(audio, 0.98)
return AudioDub(video, audio)

So I put my data with the AC3 file.

video = FFMpegSource2("d:\tmp\video.mkv")
audio = NicAC3Source("d:\tmp\audio.ac3")
Normalize(audio, 0.98)
return AudioDub(video, audio)

But when I load it on HCenc it wants to get me an m2v file, not an AC3. I thought when I loaded it there, a subprogram might open, like on Soundout.

What am I doing wrong?

[poisondeathray]

why not use Normalize() ? This will give you predictable results without clipping

Don't seem to quite get how to write this script.

I have an AC3 file that I want to normalize and still get an AC3 file output.

On the normalize() literature I found what might be an example on what I want to do.

video = AviSource("C:\video.avi")
audio = WavSource("c:\autechre.wav")
audio = Normalize(audio, 0.98)
return AudioDub(video, audio)

So I put my data with the AC3 file.

video = FFMpegSource2("d:\tmp\video.mkv")
audio = NicAC3Source("d:\tmp\audio.ac3")
Normalize(audio, 0.98)
return AudioDub(video, audio)

But when I load it on HCenc it wants to get me an m2v file, not an AC3. I thought when I loaded it there, a subprogram might open, like on Soundout.

What am I doing wrong?

HCEnc just encodes video, not audio

Add Soundout() to your script, and as soon as you play it in MPC, or vdub, or AvsP, the soundout dialog will pop up.

Note: it is not necessary to demultiplex your audio , this is another way of scripting it

FFMpegSource2("video.mkv" ,atrack=-1)
Normalize(0.98)
Soundout()

[carlmart]

HCEnc just encodes video, not audio

I know, but it also opened the soundout screen to split the audio tracks.

Add Soundout() to your script, and as soon as you play it in MPC, or vdub, or AvsP, the soundout dialog will pop up.

OK. I just did and it worked out fine. Thanks for the tip.

Note: it is not necessary to demultiplex your audio , this is another way of scripting it

FFMpegSource2("video.mkv" ,atrack=-1)
Normalize(0.98)
Soundout()

Now, THAT is interesting. I couldn't make it work yet, but it certainly looks like the way to go if it gets me an AC3 file I can use for DVD authoring. Does it?

Another interesting thing on what does normalize exactly do. I processed the AC3 file and got another AC3. So I converted it to wav just to have a look at how levels had gone up. Loaded in Soundforge and... they didn't change! Both are at the same level. Well, to be precise the highest peak in the whole new track is 0.2dB higher than the other. What does normalize do then?

[poisondeathray]

It ampliflies the entire waveform without clipping. If you used it on 5.1 audio, it uses the highest peak of any channel as the "max peak". If you go above the maximum normalize() value (instead of normalize (0.98)) you will clip (default is 1). You can use GetChannels and normalize each channel separately.

If you want the dB higher , there are probably sections where you are clipping

[carlmart]

It ampliflies the entire waveform without clipping. If you used it on 5.1 audio, it uses the highest peak of any channel as the "max peak". If you go above the maximum normalize() value (instead of normalize (0.98)) you will clip (default is 1). You can use GetChannels and normalize each channel separately.

If you want the dB higher , there are probably sections where you are clipping

What I am doing to check these levels is convert the AC3 file to wav and opening it on Soundforge 8. On older films, where there were no 5.1 remasterings, I simply adjust levels on SF up to -2dB or -3dB, checking there are no clippings anywhere along the whole track. If not I set the levels again.

When you open the file on SF, as you probably know, you can see the spectra and on it all peaks and else.

So, except if the AC3 > wav conversion is wrong, then I KNOW there are no clips on the track. And of course I am not interested in clipping anything either, as it sounds awful.

What I want is to use every bit I have left on my audio track. I have no intention of using GetChannels to adjust each channel separately: I want to adjust them all, keeping the relationships between the channels as the film was mixed.

I despise and abhor all modifications made on any film, affecting the original picture or audio, by every other people except the original creators. And I think digital and now HD remasterings provide a great chance to get to the original film.

But getting back to my original question: setting the levels as they should be, not adding noise because they are lower than they could be, is what I intend with this process. So I would like to understand better what tools I can use to get me a track that takes up every bit that's available.

[poisondeathray]

Normalize(0.98) is not the best you can do
Normalize() is the best you can do without clipping = 1.0
Normalize(1.01) would have a bit of clipping somewhere

It measures the highest peak of all channels (since you're not separating channels), to use as a reference point

Try Normalize() again and check the waveform in soundforge carefully

If you want to manually go and edit sections , you can in the audio editor as well

Or if you think you can "eyeball" it better then just use the AmplifyDB with the value you think is best

[carlmart]

Normalize(0.98) is not the best you can do
Normalize() is the best you can do without clipping = 1.0
Normalize(1.01) would have a bit of clipping somewhere

It measures the highest peak of all channels (since you're not separating channels), to use as a reference point

Try Normalize() again and check the waveform in soundforge carefully

If you want to manually go and edit sections , you can in the audio editor as well

Or if you think you can "eyeball" it better then just use the AmplifyDB with the value you think is best

OK. Please forgive my syntax ignorance on avisynth matters. But you have the learn sometime.

This is the script I am using:

video = FFMpegSource2("d:\file.mkv")
audio = NicAC3Source("d:\tmp\file.ac3")
Normalize()
Soundout

The error I get from AvsP is:

Invalid arguments to function "normalize".

What I am doing wrong?

Let me manage normalize before I go into AmplifyDB.


Edit: I corrected to "normalize(audio)" and things worked out. Sorry.

[Gavino]

Another interesting thing on what does normalize exactly do. I processed the AC3 file and got another AC3. So I converted it to wav just to have a look at how levels had gone up. Loaded in Soundforge and... they didn't change! Both are at the same level. Well, to be precise the highest peak in the whole new track is 0.2dB higher than the other.

There's something odd here.
If Normalize with 0.98 only raises the level by 0.2dB, this suggests your levels are pretty high already. This doesn't square with your original assessment of needing to raise the levels by 8.3dB.

[carlmart]

There's something odd here.
If Normalize with 0.98 only raises the level by 0.2dB, this suggests your levels are pretty high already. This doesn't square with your original assessment of needing to raise the levels by 8.3dB.

My original assessment is based on my ac3 > wav conversion, and then opening that in SF.

Just now ran another test on a different file. This time to let Soundout analize the AC3 track and see what I had.

Then I converted the AC3 file to wav with PX3convert, which I only use for synchronizing things and check dialogues. This time I wanted to see how inaccurate PX3 could be. Then I opened that wav file in SF, and to my surprise the peaks were just as on the Soundout analysis. Only it was two-track.

I tried to get a better AC3 > wav conersion using Soundout, but the wav file it got me, using Microsoft wav or Raw PCM, couldn't be opened by SF.

[volam1]

I just used audacity to convert mp3,wav to ac3 last night. the output sound is messed up.