BeHappy - AviSynth based audio transcoding tool (UPD 19-07-2006)

[fight2win]

for encoding ac3 to aac via megui, where ac3 is 192 kbps ac3 (cantonese original mono), should i go for keep original channels or convert to mono? i use force decoding via dshow, and in ac3filter, "AS Is" decoding...

[shon3i]

Just leave is as, but after encoding see what happens, is file real mono or is upmixed to stereo.

[fight2win]

please tell us proper way of encoding ac3/dts to aac, in which proper gain/volumes and all other effects are properly maintained, please...

[Doom9]

@dimzon: looking at people reporting low audio in megui I decided to have a look at the avisynthaudioencoder and compare it with what the trusty old besweet does.
I found the following differences:
besweet performs DRC. We now have DRC in nicaudio as well so I think it makes sense to use it.
besweet uses gain (be it hybrid or postgain, depending on the audio format), whereas the increase volume option in megui is a 100% normalization. I suspect we'd get better results by analyzing the samples read during the first pass, and compute a DB gain, then write a new script with the amplifydb command in it.
Do you think that's doable?

[dimzon]

Originally Posted by Doom9

I suspect we'd get better results by analyzing the samples read during the first pass, and compute a DB gain, then write a new script with the amplifydb command in it.

I believe Normalize()=="analyzing the samples read during the first pass, and compute a DB gain, then write a new script with the amplifydb command in it"

Actually I think we need to write Yet another Avisynth filter to perform non-linear DRC (like besweet's DSPguru/Ligh/Tera)

add:
this is sample http://www.ligh.de/software/booster.zip

[shon3i]

No we need only some scaner like aacgain or avsamp or foobar replay gain who will preformed RG over aac, aslo we can simulate HybridGain using first normalize than RG.

[dimzon]

Originally Posted by shon3i

No we need only some scaner like aacgain or avsamp or foobar replay gain who will preformed RG over aac, aslo we can simulate HybridGain using first normalize than RG.

Sorry, I really don't understand you!
AFAIK Normalize perform scan and compute Max peak value at first pass. Than it compute amplification factor as (factor * AbsoluteMax/FoundMax) and perform amplification on second pass...

[shon3i]

Originally Posted by dimzon

Sorry, I really don't understand you!
AFAIK Normalize perform scan and compute Max peak value at first pass. Than it compute amplification factor as (factor * AbsoluteMax/FoundMax) and perform amplification on second pass...

Just you need scaner like from foobar's ReplayGain or something like HybridGain from BeSweet, or little small app called aacgain which preform RG over aac but have same scaner like foobar (find's same db and than amplyfy it but works only with stereo).

[Doom9]

Actually, Hybridgain does the following: apply a 10dB gain at the input, encode while finding the maximum possible gain, then apply that gain as postgain after encoding.
If normalize was the same as finding and applying the maximum gain , why are the results so different in between using besweet vs behappy?

[tebasuna51]

Originally Posted by Doom9

@dimzon: looking at people reporting low audio in megui I decided to have a look at the avisynthaudioencoder

I think, like Kurtnoise, is a player problem not with encoder.

Originally Posted by Doom9

and compare it with what the trusty old besweet does.
I found the following differences:
besweet performs DRC.
We now have DRC in nicaudio as well so I think it makes sense to use it.

Of course must be a option use/not use DRC.
Boost (non-linear DRC) can be a option, but I never recommend use boost transcoding ac3 -> aac.

Originally Posted by Doom9

besweet uses gain (be it hybrid or postgain, depending on the audio format), whereas the increase volume option in megui is a 100% normalization. I suspect we'd get better results by analyzing the samples read during the first pass, and compute a DB gain, then write a new script with the amplifydb command in it.
Do you think that's doable?

AviSynth Normalize do this job (a first pass to compute a DB gain and a second pass to apply)

@shon3i ReplayGain don't work properly with ac3 5.1 like I say in this post

I make many test with AviSynth Normalize and never the aac output volume is less than ac3 input volume.

[shon3i]

Originally Posted by tebasuna51

@shon3i ReplayGain don't work properly with ac3 5.1 like I say in this post

I make many test with AviSynth Normalize and never the aac output volume is less than ac3 input volume.

That is true but when i play for example DVD via PowerDVD or other DVD players ac3 is more louder like is aac after this RG, aslo AutoGordianKnot use some Normalize over mp3 which generate very loud files, like besweet after HybridGain. Somebody told me that aac can't be used with HybridGain because aac don't have RG or post gain, both CT and Nero devs are negate that.

in this case RG works very good.

[dimzon]

Originally Posted by shon3i

That is true but when i play for example DVD via PowerDVD or other DVD players ac3 is more louder

Maybe You have Boost (non-linear DRC) option activated? Sometimes this option has some "brand" name like "Night mode" etc

[tebasuna51]

Originally Posted by shon3i

That is true but when i play for example DVD via PowerDVD or other DVD players ac3 is more louder like is aac after this RG,

This is a player problem, and can affect to many people but not to everybody. If I play ac3 with Ac3Filter configured to amplify the signal +20 dB of course I play the ac3 more loud.

Originally Posted by shon3i

aslo AutoGordianKnot use some Normalize over mp3 which generate very loud files, like besweet after HybridGain. Somebody told me that aac can't be used with HybridGain because aac don't have RG or post gain, both CT and Nero devs are negate that.

HybridGain is developed to prevent clips in encoder phase: if source is 100% (0 dB) normalized (PreGain), some samples can be encoded over 100% (a good encoder can't do this, but may exists < +2dB).
Then HybridGain is a PreGain limited (< 100%), and after a PostGain to reach the 100% (or less) in formats than accept this.

PreGain 100% is equal to HybridGain +- 2 dB, then HybridGain is more accurate than PreGain/Normalize but don't justify any volume difference. And any difference can't be previewed before the encode phase (with RG over the ac3 input)

[dimzon]

Originally Posted by tebasuna51

If I play ac3 with Ac3Filter configured to amplify the signal +20 dB

It mean You have (AmplifyDb(20) + Limiter)==Boost(Non-Linear DRC)

[shon3i]

Originally Posted by tebasuna51

If I play ac3 with Ac3Filter configured to amplify the signal +20 dB of course I play the ac3 more loud.

Yes but my ac3 filter version 1.01 plays ac3 like PowerDVD 5/6/7, settings are default, which means ac3 something about +10db louder

@all can somebody check aacgain/avsamp (AviSynth ampifler) metod because works same as RG from foobar.

[tebasuna51]

Originally Posted by dimzon

It mean You have (AmplifyDb(20) + Limiter)==Boost(Non-Linear DRC)

Nope. Is only a example.

Don't worry (be happy) with this problem. Is not a encoder problem, forget boost, non-linear DRC, ...

We need an aac 5.1 decoder, a new Bepipe (without header errors like first BeHappy) before than boost.

[dimzon]

Originally Posted by tebasuna51

We need an aac 5.1 decoder, a new Bepipe (without header errors like first BeHappy) before than boost.

Hmmm...
Why not just use DShow to decode 5.1 AAC? Actually I never transcode FROM AAC IRL...
Why do You need bePipe? Do you want to write batches?

[shon3i]

Originally Posted by dimzon

Why not just use DShow to decode 5.1 AAC? Actually I never transcode FROM AAC IRL...

Me too, but for ppl's who want back aac->ac3 or something else aac conversations will be usefull, for example i have stereo speakers and in decoder is set to dominix, so i must always disable that when i want transcode.

Originally Posted by tebasuna51

before than boost.

We not need boost, we need better first pass scener

[tebasuna51]

Originally Posted by dimzon

Hmmm...
Why not just use DShow to decode 5.1 AAC? Actually I never transcode FROM AAC IRL...

I can use DShow (I need change player settings with encoder settings and after turn to player settings), but I can't say to other people how to make that, because I don't know their DShow configuration.
You don't have problems with MEGUI DShow input?

Originally Posted by dimzon

Why do You need bePipe? Do you want to write batches?

Yeah... I like batches, not GUI's

[tebasuna51]

Originally Posted by shon3i

We not need boost, we need better first pass scener

Scanner, I suppose.
shon3i, really Normalize work fine.