Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Video Encoding > (Auto) Gordian Knot

Closed Thread
 
Thread Tools Search this Thread Display Modes
Old 24th September 2010, 11:39   #1  |  Link
temac
Registered User
 
Join Date: Apr 2006
Posts: 3
Avi Volume far too high!

I love AGK and use it quite often. But I do have an ongoing problem.

After conversion, the volume of the avi file is far too high! Is there anyway I can reduce the volume (gain) of the converted file?

Many thanks
temac is offline  
Old 24th September 2010, 14:38   #2  |  Link
Inspector.Gadget
Registered User
 
Join Date: May 2008
Posts: 1,618
If AGK has a "Normalize" option - or particularly a "Boost" option - either disable it or set the dB level to be lower.

If you're using the original DVD audio (not re-encoded to MP3), see if you can disable DRC in your decoder.
Inspector.Gadget is offline  
Old 24th September 2010, 17:10   #3  |  Link
azmoth
Registered User
 
Join Date: Dec 2009
Posts: 57
It is nothing to do with AutoGK. The normalize feature is built in and works very well, though very occasionally it can clip audio 0.5db over maximum, but not to make the audio sound like you are saying. The problem IS your source, it must be too saturated or the loud level is excessive/bass levels/dynamic range are really screwed up.

If it is too loud to start with AutoGK will only normalize it to peak levels anyway. Result: it is still loud again with 2 channels e.g if default/auto mp3 was used, or if the original audio was used it will be still too loud again!

Demux your avi/mp3 track with VirtualDubMod, use Mp3Gain and analyse it, then reduce db level to a more suitable level eg 89db/default or less. Then remux quieter track back into original. Voila solved! Google these free tools!

Last edited by azmoth; 24th September 2010 at 17:25.
azmoth is offline  
Old 27th September 2010, 10:41   #4  |  Link
temac
Registered User
 
Join Date: Apr 2006
Posts: 3
Quote:
Originally Posted by azmoth View Post
It is nothing to do with AutoGK. The normalize feature is built in and works very well, though very occasionally it can clip audio 0.5db over maximum, but not to make the audio sound like you are saying. The problem IS your source, it must be too saturated or the loud level is excessive/bass levels/dynamic range are really screwed up.
I'm sure you're right. I've been extracting the source .vob files using DVD Shrink which has very little control over the audio. Maybe I should try another DVD Ripper.
temac is offline  
Old 28th September 2010, 15:57   #5  |  Link
manono
Moderator
 
Join Date: Oct 2001
Location: Hawaii
Posts: 7,406
Neither Shrink nor any other decrypter touches the Audio. It's exactly as it is in the DVD. What kind of audio are you talking about, the original AC3 audio or reencoded MP3? What's in the AVIs that sounds so loud? If you're not sure, check an AVI in GSpot or MediaInfo, or post the AutoGK log for us to have a look.
manono is offline  
Old 29th September 2010, 05:35   #6  |  Link
yetanotherid
Banned
 
Join Date: Apr 2008
Posts: 723
What are you using to play back the AVIs? What sort of audio are you using? The original AC3 etc or are you converting to MP3? If you're using the original audio then the volume won't be touched.

AutoGK has never created an AVI with audio volume that's too high for me. I assume you're hearing distortion rather than suffering from an inability to simply turn it down?
yetanotherid is offline  
Old 29th September 2010, 10:11   #7  |  Link
temac
Registered User
 
Join Date: Apr 2006
Posts: 3
Quote:
Originally Posted by yetanotherid View Post
What are you using to play back the AVIs? What sort of audio are you using? The original AC3 etc or are you converting to MP3? If you're using the original audio then the volume won't be touched.

AutoGK has never created an AVI with audio volume that's too high for me. I assume you're hearing distortion rather than suffering from an inability to simply turn it down?
Actually I'm loading it onto a memory stick and putting it into a DVD player so I can watch it on TV. Even with the TV volume on 1, it's still pretty loud.
temac is offline  
Old 4th October 2010, 02:28   #8  |  Link
Mug Funky
interlace this!
 
Mug Funky's Avatar
 
Join Date: Jun 2003
Location: i'm in ur transfers, addin noise
Posts: 4,555
FYI, broadcast is usually peak limited to -10dBFS, and averaged either to -18 or -20. if you use a normalize function on a source that's mixed for TV, then you'll immediately be 10dB louder than broadcast level (and no doubt what your TV's volume is set for).

normalizing audio by scaling to peak should be banned for this reason - peak has no relation to actual loudness. when you're backing up DVDs (that typically are somewhat equal loudness, within a couple of dBs), but tend to use theatrical mixes or TV mixes interchangeably, all you can end up with is uneven loudness. your big budget movies will be too quiet, and your TV series will be claustrophobic and unrelenting in their assault on your ears.

if you need to boost audio for some reason, either use replaygain scans as a guide, or boost everything by a fixed amount and use peak limiting to prevent clipping. but your TV or stereo will no doubt have a volume control that can be adjusted as needed.
__________________
sucking the life out of your videos since 2004
Mug Funky is offline  
Old 5th October 2010, 20:34   #9  |  Link
setarip_old
Registered User
 
setarip_old's Avatar
 
Join Date: Aug 2005
Posts: 16,267
@temac

Hi!

You've been asked the following question twice and not answered directly (that I can see):
Quote:
What kind of audio are you talking about, the original AC3 audio or reencoded MP3?
If the audiostream of your problematic .AVI is .MP3, I believe you can load the .AVI into VirtualDub and adjust the volume...
setarip_old is offline  
Old 7th October 2010, 15:48   #10  |  Link
yetanotherid
Banned
 
Join Date: Apr 2008
Posts: 723
Quote:
Originally Posted by temac View Post
Actually I'm loading it onto a memory stick and putting it into a DVD player so I can watch it on TV. Even with the TV volume on 1, it's still pretty loud.
Check to see if your DVD player has it's own volume control. Maybe adjusting the volume there, before the TV volume, will fix it.

If you play a DVD with the same player, I assume the volume's not too high?

I've used AutoGK for years and it's volume normalising has never caused an issue, and I've played the AVIs it's created using quite a few different DVD players. Sure, some audio sounds louder than others (depending on the dynamic range), but even without normalising it's fairly rare to switch between video files and not have to adjust the volume eventually anyway.
yetanotherid is offline  
Old 7th October 2010, 15:53   #11  |  Link
azmoth
Registered User
 
Join Date: Dec 2009
Posts: 57
Quote:
Originally Posted by Mug Funky View Post
normalizing audio by scaling to peak should be banned for this reason - peak has no relation to actual loudness...
Quite correct as most of the time it usually results in clipping. Average level is far better way of normalizing a feature.
azmoth is offline  
Old 9th October 2010, 15:54   #12  |  Link
yetanotherid
Banned
 
Join Date: Apr 2008
Posts: 723
Quote:
Originally Posted by azmoth View Post
Quite correct as most of the time it usually results in clipping. Average level is far better way of normalizing a feature.
I don't suppose you'd share with us the reason behind why it results in clipping most of the time?

Assuming "average level" is a far better way of normalising, which average level do you pick? How do you know that some audio won't have enough dynamic range for it's peaks to cause clipping when normalising to the "average level" you've chosen to use? Do you normalise to your average level, then reduce it if necessary to stop peaks from clipping..... which would give you the same volume as peak normalising would have done in the first place, which of course most of the time usually results in clipping....

I'm just wondering how your "average level" method works.
yetanotherid is offline  
Old 16th October 2010, 18:03   #13  |  Link
azmoth
Registered User
 
Join Date: Dec 2009
Posts: 57
Average means exactly that: average. Using peak levels often results in overmodulation in AutoGK in either left or right channel. How it would work better is normalising to say an average of 95percent of maximum levels and avoid clipping most times. Why it clips is down to the tool used. Ask the creator!
azmoth is offline  
Old 17th October 2010, 15:59   #14  |  Link
yetanotherid
Banned
 
Join Date: Apr 2008
Posts: 723
Quote:
Originally Posted by azmoth View Post
Average means exactly that: average. Using peak levels often results in overmodulation in AutoGK in either left or right channel. How it would work better is normalising to say an average of 95percent of maximum levels and avoid clipping most times. Why it clips is down to the tool used. Ask the creator!
Well by definition normalising to an average of 95% of maximum levels.... aside from the fact it doesn't really make sense.... is really just peak level normalising that uses a slightly lower level. It's not normalising to an average level in a way I'd refer to doing so.
I assume you were trying to say you'd normalise to some average level which just happens to be a level which puts the peaks at around 95%, which brings me back to the question you didn't answer.... although I do appreciate you providing the definition of average instead. What's the average level you'd pick for normalising audio which would guarantee the peaks would only hit about 95%? Does all movie audio have the same dynamic range so you can pick an average level which will work for all of them?

I just demuxed the MP3 audio from an AVI created by AutoGK (converted from 5.1 AC3). I know AutoGK uses tools to decode the audio to a stereo wave file, it then normalises the wave file and finally that's converted to MP3. I opened the demuxed MP3 with MP3DirectCut and told it to normalise it. It found the peak level to me -1db so wouldn't alter it. I then converted the MP3 to a wave file and opened it with a wave editor. Once again the peaks were at about -1db.

So before I ask the "creator" why it clips, I think I'll need to see some evidence that it actually does. Unless of course I just happened to pick one of those AVIs which didn't contain clipped audio as it would "most of the time".

By the way, how do MP3s get "overmodulated"? I can't think of a context relating to MP3s in which that makes sense.
yetanotherid is offline  
Old 18th October 2010, 17:55   #15  |  Link
azmoth
Registered User
 
Join Date: Dec 2009
Posts: 57
Yes, that is the problem, not all have the same dynamic range yet DVD etc is supposed to have standard 5.1 track. 95% is a happy medium since actual loudness of a track varies and is a different issue. I have countlessly done encodings with AutoGK and found mp3 bitrates of 112, 160 clip while 128 or 192 do not most of the time. Don't ask me why/how overmodulation/distortion or clipping occurs technically to "T"?

However it is distortion and any tool like mp3Gain/MagixDeluxe will prove that this occurs after analysing some mp3tracks done with AutoGK so you are very wrong thinking otherwise. Why don't you lend us your extensive experience and explain why it does not occur!
azmoth is offline  
Old 19th October 2010, 04:12   #16  |  Link
yetanotherid
Banned
 
Join Date: Apr 2008
Posts: 723
Quote:
Originally Posted by azmoth View Post
95% is a happy medium since actual loudness of a track varies and is a different issue.
How can 95% be a happy medium for something which is a different issue?

Quote:
Originally Posted by azmoth View Post
I have countlessly done encodings with AutoGK and found mp3 bitrates of 112, 160 clip while 128 or 192 do not most of the time. Don't ask me why/how overmodulation/distortion or clipping occurs technically to "T"?

However it is distortion and any tool like mp3Gain/MagixDeluxe will prove that this occurs after analysing some mp3tracks done with AutoGK so you are very wrong thinking otherwise.
No I'm not wrong. It's not distortion. Not if you're using MP3Gain to detect clipping and not if, as I suspect, you're misinterpreting it's "clip track" function. Can you actually hear the distortion you claim is there? Really??

Firstly, MP3s can only be changed in volume in 1.5db steps when doing so without re-encoding. It's a limitation I don't understand fully but it likely means the values MP3Gain displays are only accurate to 1.5db anyway (I'm not saying the program is bad as I use it religiously on my music MP3s).
Secondly, MP3Gain's volume analysis and reported clipping have absolutely zero to do with whether or not the actual audio contained in the MP3 has been clipped. It's reporting whether the MP3 is likely to have peaks which will clip the preamp of the playback device after it's been ReplayGained, not whether the MP3 itself contains peaks which have been clipped. Plus MP3s can store values above 0db so they don't actually cause clipping when encoding.

It's all easy enough to confirm. Load one of your "clipping" MP3s into MP3Gain and note how it says "clip track". Now change the target volume from 89db (the default) to 79db. The clip track warning will likely disappear. Why? Mp3Gain hasn't changed the volume of the MP3 yet, but when it does all it'll be doing is reducing the volume to a point where the peaks aren't likely to clip the preamp. It hasn't magically fixed any clipped peaks in the MP3 audio itself..... it can't. Once the peaks are clipped they're clipped. All Mp3Gain can do is turn those clipped peaks up and down.

To prove it I demuxed the MP3 audio from 3 AVIs I'd recently created as they were still on the hard drive. I just picked the first 3 episodes of a series. I ran three through MP3Gain and it indicated "clip track" for all three. I checked the tags MP3Gain added to one of the AVI MP3s after running the analysis. MP3Gain was showing a clip track warning but the tag it created revealed the actual peak to be -0.988031. No clipping.

Next I loaded each one into MPDirectCut and told it to normalise them. It couldn't do so as they were too close to 0db (1.5db steps for MP3) but it reported the peaks at being -0.9db, -0.6db and -0.7db. Not one of the MP3s contained audio which exceeded 0db despite all three producing clip track warnings in MP3Gain.

Quote:
Originally Posted by azmoth View Post
Why don't you lend us your extensive experience and explain why it does not occur!
Well actually it can and it does, but it's not really anything to worry about. I probably don't understand it properly but it's to do with filtering and dither.
When converting to MP3 filtering is first applied (high pass and low pass filters)... the frequencies which are filtered are probably dependant on the bitrate being used. This filtering creates a whole new waveform to be sampled. Then dither comes into play. Dither is a kind of intentionally introduced noise which is used to randomise quantisation errors. Basically every time you alter (filter) a waveform and then resample it, because each sample can only be a certain number of values there must be "rounding". Dither randomises the way the rounding is done (for want of a better description) as random type rounding sounds natural to the human ear. So.....
It is possible the filtering and consequent dither will result in a sample which exceeds 0db but the MP3 is able to store that value faithfully so it's still not clipping the actual waveform, it's only when decoding the MP3 that the stored value greater than 0db may be clipped. If it's decoded to wave file it will, if it's simply decoded and sent to the amplifier it may not.

I say all this is nothing to worry about because it's a tiny amount of possible clipping of the waveform you're never likely to be able to hear, especially in relation to a movie soundtrack where over a two hour period the normalised wave file AutoGK created for converting to MP3 probably only hits 0db a couple of times for a fraction of a second. Probably when someone fires a gun and you're not likely to hear 0.3db of clipping there.
CD players which oversample use dithering and they can cause peaks to exceed 0db slightly for the same reason, but nobody seems to complain about it.

I'd open some of those MP3s you've run through MP3Gain and assumed were clipping, reverse any volume changes MP3Gain may have applied and then open the tags to see if any of the peak levels really exceed 0db. When you do, I think you'll probably re-evaluate your thoughts on AutoGK's normalising causing clipping.
yetanotherid is offline  
Old 19th October 2010, 12:29   #17  |  Link
azmoth
Registered User
 
Join Date: Dec 2009
Posts: 57
Originally Posted by yetanotherid

To prove it I demuxed the MP3 audio from 3 AVIs I'd recently created as they were still on the hard drive. I just picked the first 3 episodes of a series. I ran three through MP3Gain and it indicated "clip track" for all three. I checked the tags MP3Gain added to one of the AVI MP3s after running the analysis. MP3Gain was showing a clip track warning but the tag it created revealed the actual peak to be -0.988031. No clipping.

Actually it proves nothing other than that the initial analysis may be wrong. That is why you must check with other programs too and if same e.g clipping shows, then it is correct. I always find mp3Gain is fairly accurate. No, I have not re-evaluated my initial thoughts about the normalizing procedure as many tools I have used corrobate clipping effects in some tracks. I do not rely on just one tool for analysis. I agree like you write the actual distortion is nothing to worry about, but to some ears it will sound "scratchy."
azmoth is offline  
Old 19th October 2010, 12:36   #18  |  Link
Ghitulescu
Registered User
 
Ghitulescu's Avatar
 
Join Date: Mar 2009
Location: Germany
Posts: 5,769
Quote:
Originally Posted by azmoth View Post
Average means exactly that: average. Using peak levels often results in overmodulation in AutoGK in either left or right channel. How it would work better is normalising to say an average of 95percent of maximum levels and avoid clipping most times. Why it clips is down to the tool used. Ask the creator!
95% or 98% or 60% is nothing, the clipping might occur even at 50% normalisation - see the attached graph

It all depends on how well does work the analog chain (from the DAC to the speakers) in "impuls mode" (typical example cymbals and guitars).

The red dots are the digital sample, the curve is the analog output of the DAC.
Attached Images
 
Ghitulescu is offline  
Old 19th October 2010, 14:08   #19  |  Link
azmoth
Registered User
 
Join Date: Dec 2009
Posts: 57
Interesting and very informative. 95% was only a suggestion, so there is no real solution to avoid it then? Yetanotherid-indeed myself- has now no need to ask the software creator here is the tech reply for us all.
azmoth is offline  
Old 19th October 2010, 16:10   #20  |  Link
yetanotherid
Banned
 
Join Date: Apr 2008
Posts: 723
Quote:
Originally Posted by azmoth View Post
Actually it proves nothing other than that the initial analysis may be wrong. That is why you must check with other programs too and if same e.g clipping shows, then it is correct.
Err..... which part of my post explaining how I used two different programs on the same file wasn't clear enough? For the record I also used Foobar2000 to run ReplayGain on all three files to see if it gave me the same peal levels MP3Gain was writing to the MP3 tags. I'm not going to run all those test and post the information as accurate without double checking it.

Quote:
Originally Posted by azmoth View Post
I always find mp3Gain is fairly accurate.
Err.... MP3Gain was one of the programs I used.

Quote:
Originally Posted by azmoth View Post
No, I have not re-evaluated my initial thoughts about the normalizing procedure as many tools I have used corrobate clipping effects in some tracks.
Okay then. I just went to great lengths to explain how it works, and gave you examples, but even though you've obviously misunderstood how MP3Gain actually works, and I suspect you still do, feel free to believe what you want to believe.
Have you ever opened an MP3 tag and looked at the peak value MP3Gain writes to the MP3 or are you still assuming "clip track" means the waveform was clipped somewhere during the process of normalising and converted to MP3?
I don't know what "many tools" might involve or whether you understand them properly either.

Quote:
Originally Posted by azmoth View Post
I do not rely on just one tool for analysis.
Well I'd be keen to understand the point. If you run more than one program and they don't agree on how loud the peaks are then you need to work out which one is rubbish and stop using it.
Aside from ReplayGain, what other types of analysis do you run?

Quote:
Originally Posted by azmoth View Post
I agree like you write the actual distortion is nothing to worry about, but to some ears it will sound "scratchy."
Yep. Just like the thousands of people who are everyday converting normalised CDs to MP3 and sticking them on their MP3 players without changing the volume.... just like those people are constantly complaining about their MP3 selection sounding scratchy..... not.
Yet when audio in a video file could have a few tiny instances of clipping over a duration of hours, suddenly everyone's got bionic ears.
yetanotherid is offline  
Closed Thread

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 05:07.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.