MP3 audio encoder differences in Avidemux, MeGUI and Handbrake

[falco2000]

Hello. After having used Handbrake for a long time, I switched to MeGUI, and finally to Avidemux, to convert MPG (MPEG-2+MP2) movies to MP4/MKV (x264+MP3) formats.

Ignoring the fact that I always need to (manually) set an audio shift when converting with Avidemux (and I can't understand the reason why...), I'm experiencing a little problem with audio quality.

Avidemux MP3 encoder ("lame") seems to be the same used by MeGUI and Handbrake, but even using the same settings (KHz, Kbps, number of channels) the audio produced by Avidemux contains much less low frequencies than the one produced by MeGUI and Handbrake.

I also tried using a different audio format/encoder in Avidemux (AAC, etc.) but the result is still the same: the bass sounds are almost cut away.

Is there a reason for this?

Thank you very much

[hech54]

Ignoring the fact that I always need to (manually) set an audio shift when converting with Avidemux (and I can't understand the reason why...), I'm experiencing a little problem with audio quality.

I use AviDemux almost exclusively to FIX audio sync problems in videos created by other people on other editing or authoring software.
I don't use AviDemux for drastic video or audio conversions. I usually COPY the video and fix or replace audio with AviDemux.....but for the few videos I do re-encodes for....for my iPod Touch...I do notice that HandBrake's audio output is VERY nice. Not sure why.

[jagabo]

Maybe AviDemux isn't including the LFE channel in the MP3 downmix?

[jman98]

Why are you using MP3 audio? Look, I'm not here to bash you or tell you it's terrible, but it was used years ago as audio for films back in the days when Divx/Xvid was state of the art and DVD drives were too expensive, so people had to put movies on CD-R discs that they converted with Divx/Xvid. MP3 was important because it compressed a lot and it was extremely well supported because the licenses were so cheap that all the manufacturers could easily afford to pay the fees to officially and legally support it.

Today very large disk drives are cheap, DVD burners are cheap too, and there's no real need for MP3 audio. Is there really a compelling reason that you aren't just using the original audio or maybe extracting AC3 to use instead of MP3? I'm not a big fan of AAC for reasons of compatibility (it's still not as well supported as its fans would like you to think) so I get not using that. But I'm not convinced that you really and truly need to use MP3 either. I just want you to think about it.

[poisondeathray]

I'm not sure if this is the reason why, and I'm not sure about handbrake, but many of the default audio profiles in megui apply normalization (or at least they used to, I haven't used megui in a long time). Disable all the extraneous processing options in megui or match them in avidemux

[poisondeathray]

Maybe another difference is the quality setting, did you match them in megui vs. avidemux ? "it's set to "2" in avidemux by default, "0" is highest

[poisondeathray]

I also tried using a different audio format/encoder in Avidemux (AAC, etc.) but the result is still the same: the bass sounds are almost cut away.

This observation isn't as surprising, because the AAC encoder in Megui is NeroAAC or QAAC (quicktime AAC), avidemux either uses FAAC or LAV. There is no contest in quality there

If you're using updated megui and avidemux builds, both are using lame 3.99.5 , so if the settings and preprocessing are the same, then there is another problem going on

[brusno]

Is there really a compelling reason that you aren't just using the original audio or maybe extracting AC3 to use instead of MP3? I'm not a big fan of AAC for reasons of compatibility (it's still not as well supported as its fans would like you to think) so I get not using that. But I'm not convinced that you really and truly need to use MP3 either. I just want you to think about it.

I'm using AAC when I'm converting to .mp4, because I heard that the mp4 container wasn't supporting AC3 before. And that AC3 in mp4 could cause incompability with some players.

@falco2000:
Did you use VBR in all the cases?
Another option is to demux the video and audio, and process them in 2 different programs. i.e. BeSweet for the audio conversion. Then you might have more options to choose between (quality parameters)

[jman98]

I'm using AAC when I'm converting to .mp4, because I heard that the mp4 container wasn't supporting AC3 before. And that AC3 in mp4 could cause incompability with some players.

The idiots who wrote the MP4 spec (probably Apple but I truly don't know) did not at first support AC3 audio. That is true. But it's been years since they changed that to support it. I guess you did not think about just testing it to see if it works.

I think it's very foolish to let a piece of hardware dictate that you convert everything. It would be better to buy a better playback device, like a Western Digital or similar media streamer, than to have to use MP3 because some old crappy player requires it when it should allow AC3.

I don't know what your playback device is, but could you use MKV instead of MP4 containers? Devices that support MKV are generally less fussy than those that support MP4 and H.264 video + AC3 audio should be fine anything that support MKV.

[brusno]

The idiots who wrote the MP4 spec (probably Apple but I truly don't know) did not at first support AC3 audio. That is true. But it's been years since they changed that to support it. I guess you did not think about just testing it to see if it works.

I think it's very foolish to let a piece of hardware dictate that you convert everything. It would be better to buy a better playback device, like a Western Digital or similar media streamer, than to have to use MP3 because some old crappy player requires it when it should allow AC3.

I don't know what your playback device is, but could you use MKV instead of MP4 containers? Devices that support MKV are generally less fussy than those that support MP4 and H.264 video + AC3 audio should be fine anything that support MKV.

I convert to .mp4 with AAC audio mainly because I wanted to watch the video on my iPad and via AppleTV. This is because the iPad (Apple) doesn't support AC3 sound (and that's too bad). However, I made a test with AC3 sound in mp4 container now, and it played just fine on my stand alone blu-ray player
The iPad accepts .mp3 also by the way.

PS: I see it's written a lot about AC3 and licensing on the internet. Maybe AC3 isn't a safe bet in the future because it's expensive to integrate in players?

[jagabo]

PS: I see it's written a lot about AC3 and licensing on the internet. Maybe AC3 isn't a safe bet in the future because it's expensive to integrate in players?

It' can't be all that expensive. Every US$15 DVD player plays it. AC3 will be around for a long time because it's part of the DVD spec. It's part of the Blu-ray spec too. AC3 patents must be nearing their expiration date as the first AC3 products appeared in ~1995.

[hello_hello]

I'm not a big fan of AAC for reasons of compatibility (it's still not as well supported as its fans would like you to think) so I get not using that. But I'm not convinced that you really and truly need to use MP3 either. I just want you to think about it.

I think it's very foolish to let a piece of hardware dictate that you convert everything. It would be better to buy a better playback device, like a Western Digital or similar media streamer, than to have to use MP3 because some old crappy player requires it when it should allow AC3.

Mmmmmm...... sounds like one theory for one audio format, and another theory for the other.

[hello_hello]

The only thing I could image effecting the presence of low frequencies after downmixing and converting to MP3 would be the inclusion (or not) of the LFE channel. Or maybe if the AC3 dynamic range compression is being applied by one program and not the other. Given you get the same result when downmixing and converting to AAC, you'd assume it's do with the downmixing and nothing to do with the actual encoding as such.
When downmixing, the LFE channel is actually optional. It only includes a baas "boost" but contains no additional information. In fact the Dolby AC3 spec doesn't include the LFE channel when a player downmixes multichannel audio to stereo. Are you hearing a general lack of low frequencies all the time, ie even when the audio is mainly speech etc?
I just checked MeGUI's log file after running a quick test encode and it does include the LFE channel. Can you extract log files form HandBrake/Avidemux?

Are you sure it's not that one program is normalizing the audio while the other isn't? It's possible a volume difference may make the low frequencies sound different.

Personally I use AAC for most conversions as it's far more widely supported than it's detractors would have you believe (probably more widely than DTS), but AAC, MP3, AC3..... it really doesn't matter, as at a high enough bitrate pretty much any format/encoder is going to give you a transparent encode, but I do tend to keep the original AC3 audio myself as file size isn't too much of an issue for me. I generally convert DTS to multichannel AAC as that can save you around 1GB per movie, but each to their own.....

[falco2000]

Maybe AviDemux isn't including the LFE channel in the MP3 downmix?

The source movie has a simple stereo audio stream. No LFE channel, as far as I know.

Why are you using MP3 audio?

Because, when I started my adventure in the conversion of movies months ago, I tried AAC in Handbrake and the resulting audio had much less high frequencies than MP3 using the same settings. I couldn't find the way to fix it, so I chose MP3.

many of the default audio profiles in megui apply normalization (or at least they used to, I haven't used megui in a long time). Disable all the extraneous processing options [...]

I didn't see any extra processing functions enabled in the three mentioned programs.

Maybe another difference is the quality setting, did you match them in megui vs. avidemux ? "it's set to "2" in avidemux by default, "0" is highest

Good suggestion: I had already tried to set quality to "1" (getting no differences) since I didn't think to a possible "0" value, but I tried it and it's the same: no audible difference.

If you're using updated megui and avidemux builds, both are using lame 3.99.5 , so if the settings and preprocessing are the same, then there is another problem going on

I think so too.

Did you use VBR in all the cases?

In none of them.

Another option is to demux the video and audio, and process them in 2 different programs.

That's what I wanted to avoid: I'd like to use a program like Avidemux that makes the whole conversion in a single process, since I obtained audio-video sync loss in my cleaned-from-commercials edited+converted movies even using MeGUI.

I don't know what your playback device is, but could you use MKV instead of MP4 containers? Devices that support MKV are generally less fussy than those that support MP4 and H.264 video + AC3 audio should be fine anything that support MKV.

The container is not essential to me, I'll test it, thank you.

I just checked MeGUI's log file after running a quick test encode and it does include the LFE channel. Can you extract log files form HandBrake/Avidemux?

Just to save time, I'll check it only if the LFE channel is something that could be included in a stereo audio (2 channels) without being explicitly mentioned in the movie properties (using MediaInfo, for example). Is it so? Because, as I hinted at above, MediaInfo tells that the source movie has a simple stereo audio stream (2 channels).

Thank you very much for your answers.

[Hoser Rob]

I'm with jman here. Why are you converting the audio at all? Unless there's a playback device issue there's no reason to. I always play video from the computer so there's no problem. Any decent software can handle that stuff.

[falco2000]

I'm with jman here. Why are you converting the audio at all?

Because I'm storing small versions of my DVB-T recordings. I need the minimum acceptable audio quality, that's 64Kbps in my case. But I'm talking about "64Kbps using Handbrake or MeGUI". The problem is that, using Avidemux, low frequencies are unexpectedly lost, while are perfectly audible in Handbrake and MeGUI outputs.

[falco2000]

I did some tests and I got news that I'll post as soon as possible. Please don't reply in the meanwhile.

[falco2000]

Ok, I tested the following Avidemux outputs, with audio always encoded at 64Kbps (usually the only available parameter for all of the used encoders, except for "MP3 lame" for which I also chose "CBR" and "quality: 0". "Disable reservoir" did no difference):

Best result:

Avidemux test movie AAC (FAAC).avi
Avidemux test movie AAC (FAAC).mkv
Avidemux test movie AAC (FAAC).mp4
Avidemux test movie AAC (FAAC) mp4v2.mp4

Really very similar to the previous one but just slightly worse:

Handbrake test movie MP3 (lame).mp4

Slightly worse than the previous one:

Avidemux test movie AAC (LAV).avi
Avidemux test movie AAC (LAV).mkv
Avidemux test movie AAC (LAV).mp4
Avidemux test movie AAC (LAV) mp4v2.mp4

Audibly worse than AAC and equal between them (i.e.: Aften and LAV):

Avidemux test movie AC3 (aften).avi
Avidemux test movie AC3 (aften).mkv
Avidemux test movie AC3 (aften) mp4v2.mp4

Avidemux test movie AC3 (LAV).avi
Avidemux test movie AC3 (LAV).mkv
Avidemux test movie AC3 (LAV) mp4v2.mp4

Similar "final" score as AC3 but: as a pro, MP3 kept more high frequencies, as a con it has the typical phaser-like low quality MP3 effect on high frequencies:

Avidemux test movie MP3 (lame).avi
Avidemux test movie MP3 (lame).mkv
Avidemux test movie MP3 (lame).mp4
Avidemux test movie MP3 (lame) mp4v2.mp4

Worse than the previous one:

Avidemux test movie MP2 (twolame).avi
Avidemux test movie MP2 (twolame).mkv
Avidemux test movie MP2 (twolame).mp4
Avidemux test movie MP2 (twolame) mp4v2.mp4

Worst of all, because it lost high frequencies if compared to the previous one:

Avidemux test movie MP2 (LAV).avi
Avidemux test movie MP2 (LAV).mkv
Avidemux test movie MP2 (LAV).mp4
Avidemux test movie MP2 (LAV) mp4v2.mp4

I used four different containers when possible, just to be sure.

So I got the evidence for what I wrote in my first post: "Handbrake MP3 Lame audio" is much more "powerful" than "Avidemux MP3 Lame one". It keeps more low frequencies and more "presence". I don't understand the reason why, since the encoder is the same and apparently no extra settings had been enabled.

The difference between the "winner", AAC FAAC output and Handbrake's MP3 Lame, is that the latter has a slight phaser-like effect in the high frequencies.

I think that I'll go for AAC FAAC from now on, since I don't need to play these movie on particular devices. But I'd still like to understand the reason for the difference between the two MP3 Lame outputs.

[jagabo]

How about some samples?

[brusno]

You may also try BeSweet option in StaxRip, and compare with your previous test.

[falco2000]

Ok, I got it. Avidemux creates a stereo audio stream, while Handbrake creates a joint stereo one (info by MediaInfo; although Handbrake for Windows doesn't mention it when configuring audio settings, it simply shows "mixdown: stereo"). That's the reason, AFAIK, that explains the resulting higher quality of one of the two movies.

Now the question is: how could I force Avidemux MP3 Lame encoder to create a joint stereo output instead of the stereo one?

[jagabo]

You would certainly want joint stereo (or mono) with MP3 at 64 kbps.

[falco2000]

You would certainly want joint stereo (or mono) with MP3 at 64 kbps.

Yes, but I still don't see the way to do it.

Moreover, I also found out that the probable reason for the higher quality audio of Avidemux's AAC (faac) encoding compared to the Handbrake's joint stereo MP3 one is that the output has an average bitrate of about 100Kbps (info by MediaInfo) and not 64Kbps like I selected in Avidemux audio encoder configuration. I tried to set 56Kbps too, but the output is always 100Kbps.

[jagabo]

You would certainly want joint stereo (or mono) with MP3 at 64 kbps.

Yes, but I still don't see the way to do it.

I saw an option for a mono downmix but no control over stereo/joint stereo MP3 encoding.

Moreover, I also found out that the probable reason for the higher quality audio of Avidemux's AAC (faac) encoding compared to the Handbrake's joint stereo MP3 one is that the output has an average bitrate of about 100Kbps (info by MediaInfo) and not 64Kbps like I selected in Avidemux audio encoder configuration. I tried to set 56Kbps too, but the output is always 100Kbps.

I wouldn't trust what MediaInfo tells you. It only looks at what's flagged in the header. That may be peak kbps or something else. Demux the audio tracks and compare their size. Or use a player that shows you the realtime bitrate as the audio is being played. In general, AAC is superior to MP3, especially at low bitrates.

[hello_hello]

Were they both producing the same type of AAC audio? ie are they AAC-LC or HE-AAC etc. One GUI may automatically change encoding methods according to the bitrate which would probably also effect the quality (I'd assume improve it). I've not experimented with the different flavours of AAC myself as I always encode using the NeroAAC encoder's quality mode (variable bitrate) so it's always AAC-LC.

I haven't looked back through the thread to refresh my memory, but is it really necessary to use such a low bitrate? If you do the math (and mine is correct), at 128kb/s an hour of stereo audio will be roughly 56MB in size. Therefore at 64kb/s it'll be around 28MB per hour. 128kb/s should give you fairly decent quality using MP3, and better if it's AAC, so is it worth risking noticeably reducing the audio quality to save 28MB of file size/disc space per hour of video?

[falco2000]

Demux the audio tracks and compare their size.

Thanks for the suggestion Jagabo, I'll check it.

I haven't looked back through the thread to refresh my memory, but is it really necessary to use such a low bitrate? If you do the math (and mine is correct), at 128kb/s an hour of stereo audio will be roughly 56MB in size. Therefore at 64kb/s it'll be around 28MB per hour. 128kb/s should give you fairly decent quality using MP3, and better if it's AAC, so is it worth risking noticeably reducing the audio quality to save 28MB of file size/disc space per hour of video?

I own a 1000GB USB HDD and I already archived 90GB of "micro-ed" tv series episodes that I've already watched in original resolution and don't plan to watch them again except for extraordinary reasons. They are approximately 1400 files up to now, and the current remaining 910GBs are used for new recordings or for full resolution compressed movies (already watched but worthy of a second watch in the future, like very good or famous movies).

I can't go below 320x240 video + 64Kbps audio for the "micro-ed" movies, this is my best compromise between "watchability" and compression).

Maybe one day I'll let go this habit of keeping these micro versions and I'll immediately delete everything I've watched unless it's a real masterpiece.

[falco2000]

Demux the audio tracks and compare their size.

Thanks for the suggestion Jagabo, I'll check it.

I did it. MediaInfo was right: for the same piece of audio encoded by Avidemux at "nominal" 64Kbps, AAC size is about 184KB, while MP3 is only 113KB. This confirms that even if I select AAC 64Kbps in Avidemux, the encoding is actually done at about 98Kbps. It seems an Avidemux "bug".

But the other fact that I noticed is that, while AAC 184KB audio file sounds audibly much better than MP3 joint-stereo 113KB one, that's not true anymore if I increase the MP3 to about the same bitrate of the AAC: that means that I selected the closer bitrate available in Avidemux to obtain an MP3 file with about the same size of the AAC, and this bitrate is 96Kbps (below there's 80 and above there's 128: there's no textfield to set a custom value).

So I got 184KB (AAC at 98Kbps) vs. 170KB (MP3-JS at 96Kbps). Now the quality is very similar according to my ears: AAC output is only slightly better than MP3 output. And maybe if I increased MP3-JS to 98Kbps to have two same-size files, I may couldn't even notice any difference at all.