eac3to - audio conversion tool

[hubblec4]

There are more differences. Eac3to will not really work with UHD-BD at the moment.
https://forum.doom9.org/showthread.php?p=1810285#post1810285

[colpoz]

Is there any way to extract only the subtitles flagged as forced instead of the entire track??

Example if the English track contains "1000 normal and 50 forced captions" I want to extract only the 50 forced ones.

Is it possible using only eac3to command line?

[LigH]

Probably not (at least I see no option mentioned in its help output). The purpose of eac3to is mainly to handle audio. You may need a tool specialized in handling subtitles instead.

[stax76]

Is there any way to extract only the subtitles flagged as forced instead of the entire track??

Example if the English track contains "1000 normal and 50 forced captions" I want to extract only the 50 forced ones.

Is it possible using only eac3to command line?

As far as I know such disks are extremely rare, what's the name of the title? Tools you can try are BDSup2Sub(++) and Subtitle Edit.

[Q-the-STORM]

As far as I know such disks are extremely rare, what's the name of the title? Tools you can try are BDSup2Sub(++) and Subtitle Edit.

I've encountered a lot of those, it's not that rare... you can extract them with BDSub2Sub... just drop the file in there, navigate to "File" -> "Save/Export" -> tick "Export only Forced" -> click on "save"...

If you prefer working in CLI, use --forced-only

[stax76]

I've encountered a lot of those, it's not that rare... you can extract them with BDSub2Sub... just drop the file in there, navigate to "File" -> "Save/Export" -> tick "Export only Forced" -> click on "save"

Can you name same titles, I'm always looking to improve my sample collection.

[von Suppé]

Take into account that forced subtitles can be flagged as "forced" in a total subtitle stream,
but they can also be muxed / authored as a separate stream.

[stax76]

I'm interested in titles that have it flagged as forced, ideally with spoken dialog and small file size.

[nevcairiel]

Most of the Avatar Blu-ray releases use flagged forced frames for alien dialog, sometimes even with normal frames active at the same time, which is a good test if the separation works properly.

[stax76]

It's a bit long, at least the price is low.

[nevcairiel]

Blu-rays are the only sources of commercial PGS subtitles, what do you expect? :p

[stax76]

Probably unrealistic to find but something smaller like 100 min, 30 GB and decent movie I've not already seen.

[Q-the-STORM]

I don't remember which ones, it's not like I wrote it down... Definitely a few seasons of True Blood have them (not all, some have separate forced subs), same goes for Game of Thrones...

[thomaz909]

When i use "-normalize" it creates wav files only to get the maxiumum gain value of the source.
Isn't it possible to just read the gain value of the source instead of doing a complete decode to wav (some ac3 640 sources result in very huge 64 bit float wavs)?

[LigH]

Gain values in headers are no guarantee to avoid an integer overflow or a suboptimal volume.

[Boulder]

A good option would be to output the first pass to NUL if normalization is forced by the user.

[tebasuna51]

Maybe, but with a wav file the second pass don't need re-decode source.

[thomaz909]

isn't encoding the bottleneck in the whole process?

i would really love to have an option to -normalize "onthefly".
ssd wear, small ramdisks.
~107mins ac3 640 produced a 14gb 64bit wav 5.1 file, only to get the 4-5 digit db value.

[Asmodian]

Maybe, but with a wav file the second pass don't need re-decode source.

Or even worse, re-extract if from the BD!

isn't encoding the bottleneck in the whole process?

i would really love to have an option to -normalize "onthefly".
ssd wear, small ramdisks.
~107mins ac3 640 produced a 14gb 64bit wav 5.1 file, only to get the 4-5 digit db value.

To do normalizing well you need to read the entire file before writing anything. Maybe it could be cached in memory (assuming there is enough free memory) or recompressed to flac for the intermediate file, to reduce IO, but true on the fly normalizing is a bad idea.

[sneaker_ger]

To do normalizing well you need to read the entire file before writing anything.
But you don't need to store the decoded data of the first pass. Only store the peak volume. That's just one value that gets updated when you find a new peak. O(1).

[tebasuna51]

~107mins ac3 640 produced a 14gb 64bit wav 5.1 file, only to get the 4-5 digit db value.

BTW, -normalize is only recomended when downmix, don't use normalize when recode 5.1 -> 5.1

[Boulder]

BTW, -normalize is only recomended when downmix, don't use normalize when recode 5.1 -> 5.1Why not? Is it due to the dynamics between the channels?

[mstrong]

I have learned how to use the command line version of eac3to. The process creates a gap file. How do I mux my files back together using the gap file?

How do I do it using eac3to? I am also wondering if it can be done using MKVToolNix GUI.

[tebasuna51]

Why not?

Well, you can do it if you listen low volume, but most the times the input source (from BD's) have a correct volume level without maximize.

Is it due to the dynamics between the channels?

If you use eac3to to decode (ignoring any Dynamic Range Compression in source) and a free AC3 encoder (Aften or ffmpeg) than don't add DRC to output by default, the dynamics between channels are preserved, because the same gain is applied over all channels.

Only if you use a Dolby Digital encoder, than create DRC, the normalize can distort the initial balance between channels.

[tebasuna51]

I have learned how to use the command line version of eac3to. The process creates a gap file. How do I mux my files back together using the gap file?

How do I do it using eac3to? I am also wondering if it can be done using MKVToolNix GUI.

When eac3to create a gap file it do, automatically, a second pass to correct the detected problems, and the corrected files are ready to mux (if all work fine of course).

AFAIK only eac3to can use that gap file.

[Boulder]

Well, you can do it if you listen low volume, but most the times the input source (from BD's) have a correct volume level without maximize.
What I've noticed is that many older movies have a quite big headroom (several dB's) even in the 5.1-channel lossless track. The mono and stereo tracks are all quite low in volume, sometimes eac3to shows that you need to amplify by almost 10dB to normalize. That's why I've always enabled normalization apart from encoding the original audio track to FLAC to save some space.

[Q-the-STORM]

How do I mux my files back together using the gap file?
Just run the same command again. When eac3to sees that there is a gap file, it will use it.

[hello_hello]

isn't encoding the bottleneck in the whole process?

i would really love to have an option to -normalize "onthefly".
ssd wear, small ramdisks.
~107mins ac3 640 produced a 14gb 64bit wav 5.1 file, only to get the 4-5 digit db value.

You could effectively do it after a ReplayGain or EBU R128 scan. The audio still needs to be scanned first, but it should be faster than converting to wave. Once you have the peak values, you know how to adjust the volume when converting.

My old quad-core can scan a 2 hour, 5.1ch AC3 file in around 40 seconds (oversampling disabled), but it can also scan four of them simultaneously in about 40 seconds. I use foobar2000 for that sort of thing as it's ReplayGain scanner is quite fast.

It can also adjust the level to the ReplayGain target volume when converting, but you can specify an additional increase or decrease, so by enabling the "apply gain and prevent clipping" option with a 20dB increase, it'll limit the increase to prevent clipping according to the ReplayGain info, and the end result is peak normalising.
Adjusting to the ReplayGain target volume minus 5dB gives you the standard EBU R128 target volume, which is how I'd normalise after downmixing related files to stereo (episodes of a TV show etc), although it doesn't hurt to scan the output files and check the peaks just to be sure.

Why has ReplayGain/R128 scanning been almost completely ignored by programs related to video conversion?

Avisynth runs two passes, but does it write a wave file? If not that'd be another way to normalise without writing an intermediate file.

What I've noticed is that many older movies have a quite big headroom (several dB's) even in the 5.1-channel lossless track. The mono and stereo tracks are all quite low in volume, sometimes eac3to shows that you need to amplify by almost 10dB to normalize. That's why I've always enabled normalization apart from encoding the original audio track to FLAC to save some space.

That's no doubt true but it probably also means they have less dynamic range than newer movies, but the "average" volume, or at least the dialogue volume, would probably be very similar. Well that's my theory....

[MrVideo]

For the life of me, I have never been able to understand why anyone would want to screw with the volume of a movie/TV show soundtrack. The mix is what the original director/producers/audio engineers wanted it to be in their product. Why do some think that they know better than those who put it together in the first place?

[nevcairiel]

For the life of me, I have never been able to understand why anyone would want to screw with the volume of a movie/TV show soundtrack. The mix is what the original director/producers/audio engineers wanted it to be in their product. Why do some think that they know better than those who put it together in the first place?

Because between the producers and you getting it, there were so many different people messing with it, that any original intent is already lost, especially when handling broadcast, to a lesser extend with optical media - and most people would otherwise just reach for the remote to adjust the volume.
These days most broadcast content is already volume leveled with EBUR128 though, since a lot of stations adopted this standard.

Note that volume leveling is only an overall attenuation of the volume. It does not change its dynamic range, or anything like that, it just changes the overall volume by a constant factor, just like you would change the volume on your speakers.
Thats why a 2-pass process is needed, analyze the full track to find its volume level, and then attenuate by that level.

[tebasuna51]

From the beginning eac3to was designed to transcode audio without lose quality of source.

And only a Peak Normalization (https://en.wikipedia.org/wiki/Audio_normalization) is allowed with -normalize.
Of course you can limit the peaks to avoid excesive volume, but the dynamic range is preserved.

A Loudness Normalization based in RMS values instead peak values, like ReplayGain and EBU R128 do, is not always possible without lose dynamic range.
Of course any user can do this with other tools than eac3to.

[hello_hello]

For the life of me, I have never been able to understand why anyone would want to screw with the volume of a movie/TV show soundtrack. The mix is what the original director/producers/audio engineers wanted it to be in their product. Why do some think that they know better than those who put it together in the first place?

If there's dynamic range reduction info saved to AC3 audio, it would have been the director/producers/audio engineers who put it there. You can decode without using that information to reduce the dynamic range, but it's part of the AC3 spec.

I generally compress audio for movies/TV shows I copy to the hard drive connected to the TV, but that's only because I can do it better than the TVs "night mode", and it's only a temporary copy for viewing in the small hours.

I don't think the recent posts referred to compressing the audio though, only adjusting the level up or down as a whole, or adjusting it so the peaks are near maximum when encoding.

[Ripman]

As far as I know such disks are extremely rare, what's the name of the title? Tools you can try are BDSup2Sub(++) and Subtitle Edit.

MakeMKV shows when sub are forced, and if they're not empty, the forced subs are left in the resultant mkv.

Examples of BDs with forced subs that MakeMKV sees, and that can be extracted with mkvextract or Subtitle Edit or AegisSub etc.: Star Wars VI Return of the Jedi; Patton; Kill Bill BDs; Tora Tora Tora; The Walk; The Martian.

[mstrong]

Just run the same command again. When eac3to sees that there is a gap file, it will use it.

I am assuming the gap file needs to be in the same folder as the video and audio source files?

Could you give me an example syntax for muxing a video and audio file into the same MKV?

Thank you.

[Q-the-STORM]

I am assuming the gap file needs to be in the same folder as the video and audio source files?
Yes, the gap file is being created in the correct folder anyways, so there's no need to move anything, just run the command a second time.

Could you give me an example syntax for muxing a video and audio file into the same MKV?
Can't be done with eac3to, you gotta use mkvmerge for that...

[thomaz909]

i just want to "normalize" to 0db.
that means, it scans all channels, takes the highest overall gain found and then amplifies all channels by this value.
for example L=-5db, R=-6db, C=-2db, SL=-12db, SR=-13db, then the center channel has the highest gain with -2db and finally all channels get an amplification by +2db.
no drc, compressor, loudness maximizer, limiter, etc.
and that all in the most efficient, resource saving way. :)

[tebasuna51]

i just want to "normalize" to 0db.
that means, it scans all channels, takes the highest overall gain found and then amplifies all channels by this value.
for example L=-5db, R=-6db, C=-2db, SL=-12db, SR=-13db, then the center channel has the highest gain with -2db and finally all channels get an amplification by +2db.
no drc, compressor, loudness maximizer, limiter, etc.
and that all in the most efficient, resource saving way. :)

That is just like eac3to -normalize work.

To be clear is a "Peak Normalization" than don't change dynamics, and not a "Loudness Normalization (RMS)"

[MrVideo]

i just want to "normalize" to 0db.
that means, it scans all channels, takes the highest overall gain found and then amplifies all channels by this value.
for example L=-5db, R=-6db, C=-2db, SL=-12db, SR=-13db, then the center channel has the highest gain with -2db and finally all channels get an amplification by +2db.
no drc, compressor, loudness maximizer, limiter, etc.
and that all in the most efficient, resource saving way. :)
I hate to say this, but this a a bad example, because I dare you, in a blind test, that you will not be able to notice that 2dB difference.

[thomaz909]

it would be nice if eac3to ignores -downXX switch when source is lower than the specified bitrate.
for example decoding a 64bit ac3 with -down32 then it "downbits" to 32bit.
when decoding a 24bit dts it ignores the switch and decodes to 24bit instead of "upbitting" to 32bit.

or add -downXXXX (for example -down6432)
then it only downs 64bit to 32bit and leaves all other input bitrates untouched.

[sneaker_ger]

It already works like this, I believe. Can you provide any example (sample file + log) where it doesn't?

[tebasuna51]

it would be nice if eac3to ignores -downXX switch when source is lower than the specified bitrate.

Maybe you mistake "bitrate" with "bitdepth"?

-downXX fix the bitdepth of decoded PCM data.

for example decoding a 64bit ac3 with -down32 then it "downbits" to 32bit.

Is possible to have a 64 Kb/s AC3 (I never see), but AC3 don't have any bitdepth then 64bit AC3 don't have sense.

when decoding a 24bit dts it ignores the switch and decodes to 24bit instead of "upbitting" to 32bit.

- If you are talking about a DTS-MA 24 bits, for what you want 32 bits with the last 8 bits all to "0"?

eac3to try to be smart and ignore usseless parametres, even check if last 8 significant bits are all to "0" to output only 16 bits.

- If you are talking about a standard DTS, like AC3, don't have bitdepth. Does not exist a standard DTS 24 bits.

When eac3to decode a lossy format (AC3, DTS, MP3, ...) internally work with 64 bits float samples and, at output by default, downsample to 24 bits int because is enough precission.

If you want preserve the 64 bit float you can add the parameter:

-full

[stax76]

@madshi

I would like to request 0.1 intervals for the percentage status display/output.

Some years ago I suggested to use an alternative or additional mkv muxer because many people don't like to install directshow filters, configuring DirectShow and in general configuring and repairing Windows requires expertise many people don't have.

This summer I received a email with following content:

Shit don’t work and it sucks.

I was mostly amused about it, two days ago I received something more concrete from the same kind person:

So explain to me why is it you made 100 versions crappy with the haali filter, then I am telling everyone not to use it because of that stupid haali filter and now I see version 101 does not have it? You made me look like a fool, you are a pile of shit and I will love you track and hunt you down for literally being the shittiest developer on planet earth.

I don't know who this person is and these two mails is the only thing I ever heard from him, I do not reply such mails.

In staxrip Haali is optional even for eac3to usage and I didn't remove or change anything. For VC1 I think eac3to users need Haali because I don't think there is a avs/vs source filter that can handle raw VC1.

While staxrip supports many different tools and workflows eac3to is still popular and I always ensured it has the best possible support.

[Music Fan]

Is there a way to select the output folder when using the -demux function ?

[tebasuna51]

You can try:

eac3to "input" "D:\tmp\output.*"

[Music Fan]

Thanks, good trick !

[hubblec4]

Hi madshi

Exists a hidden switch to get the PID's for the streams? The order you display the streams is different to the mpls and the m2ts.

[Snowknight26]

-logpids

[hubblec4]

many thanks for the info

[stax76]

@madshi

There seem to be a problem with a AC3 file, it was reported by a staxrip user so I don't have the source but if you want to work on it I might be able to get the file for you.

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Source file MediaInfo
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

F:\Movie.mkv

General
Complete name : F:\Movie.mkv
Format : Matroska
Format version : Version 1
File size : 1.46 GiB
Duration : 56 min 46 s
Overall bit rate : 3 677 kb/s
Encoded date : UTC 2009-06-22 03:44:10
Writing application : mkvmerge v2.4.0 ('Fumbling Towards Ecstasy') built on Nov 7 2008 17:24:23
Writing library : libebml v0.7.8 + libmatroska v0.8.1

Video
ID : 1
Format : AVC
Format/Info : Advanced Video Codec
Format profile : High@L3.1
Format, CABAC : Yes
Format, ReFrames : 3 frames
Codec ID : V_MPEG4/ISO/AVC
Duration : 56 min 46 s
Bit rate : 3 289 kb/s
Width : 1 280 pixels
Height : 720 pixels
Display aspect ratio : 16:9
Frame rate mode : Constant
Frame rate : 23.976 (24000/1001) FPS
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Scan type : Progressive
Bits/(Pixel*Frame) : 0.149
Stream size : 1.28 GiB (88%)
Writing library : x264 core 67 r1162M f7bfcfa
Language : English
Default : Yes
Forced : No

Audio
ID : 2
Format : AC-3
Format/Info : Audio Coding 3
Format, Endianness : Big
Codec ID : A_AC3
Duration : 56 min 46 s
Bit rate mode : Constant
Bit rate : 384 kb/s
Channel(s) : 2 channels
Channel positions : Front: L R
Sampling rate : 48.0 kHz
Frame rate : 31.250 FPS (1536 spf)
Bit depth : 16 bits
Compression mode : Lossy
Stream size : 156 MiB (10%)
Service kind : Complete Main
Default : Yes
Forced : No

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Demux mkv using mkvextract 13.0.0 x64
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

C:\StaxRip\Apps\MKVToolNix\mkvextract.exe tracks F:\Movie.mkv 1:"F:\Movie_temp\Movie ID1.ac3" --ui-language en

Extracting track 1 with the CodecID 'A_AC3' to the file 'F:\Movie_temp\Movie ID1.ac3'. Container format: Dolby Digital (AC-3)

Start: 9:08:04 PM
End: 9:08:07 PM
Duration: 00:00:02

General
Complete name : F:\Movie_temp\Movie ID1.ac3
Format : AC-3
Format/Info : Audio Coding 3
File size : 156 MiB
Duration : 56 min 46 s
Overall bit rate mode : Constant
Overall bit rate : 384 kb/s

Audio
Format : AC-3
Format/Info : Audio Coding 3
Format, Endianness : Big
Duration : 56 min 46 s
Bit rate mode : Constant
Bit rate : 384 kb/s
Channel(s) : 2 channels
Channel positions : Front: L R
Sampling rate : 48.0 kHz
Frame rate : 31.250 FPS (1536 spf)
Bit depth : 16 bits
Compression mode : Lossy
Stream size : 156 MiB (100%)
Service kind : Complete Main

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Audio Source File MediaInfo
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

General
Complete name : F:\Movie_temp\Movie ID1.ac3
Format : AC-3
Format/Info : Audio Coding 3
File size : 156 MiB
Duration : 56 min 46 s
Overall bit rate mode : Constant
Overall bit rate : 384 kb/s

Audio
Format : AC-3
Format/Info : Audio Coding 3
Format, Endianness : Big
Duration : 56 min 46 s
Bit rate mode : Constant
Bit rate : 384 kb/s
Channel(s) : 2 channels
Channel positions : Front: L R
Sampling rate : 48.0 kHz
Frame rate : 31.250 FPS (1536 spf)
Bit depth : 16 bits
Compression mode : Lossy
Stream size : 156 MiB (100%)
Service kind : Complete Main

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Audio encoding using eac3to 3.31 x86
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

C:\StaxRip\Apps\eac3to\eac3to.exe "F:\Movie_temp\Movie ID1.ac3" "F:\Movie_temp\Movie ID1_a1.m4a"
-quality=0.23 -normalize -downStereo -progressnumbers

AC3, 2.0 channels, 0:56:46, 384kbps, 48kHz, dialnorm: -27dB
Removing AC3 dialog normalization...
Decoding with libav/ffmpeg...
Writing WAV...
Creating file "F:\Movie_temp\Movie ID1_a1.m4a.pass1.wav"...
Caution: The WAV file is bigger than 4GB.
Some WAV readers might not be able to handle this file correctly.
Starting 2nd pass...
Reading WAV...
Reducing depth from 64 to 32 bits...
Encoding AAC <0.23> with NeroAacEnc...
Applying 0.95dB gain...
The original audio track has a constant bit depth of 64 bits.
The processed audio track has a constant bit depth of 32 bits.
eac3to processing took 3 minutes, 34 seconds.
Done.

Start: 9:08:08 PM
End: 9:11:43 PM
Duration: 00:03:34

General
Complete name : F:\Movie_temp\Movie ID1_a1.m4a
Format : MPEG-4
Format profile : Base Media / Version 2
Codec ID : mp42 (M4A /mp42/isom)
File size : 76.3 MiB
Duration : 2 h 48 min
Overall bit rate mode : Variable
Overall bit rate : 63.4 kb/s
Encoded date : UTC 2017-09-22 01:08:31
Tagged date : UTC 2017-09-22 01:11:42

Audio
ID : 1
Format : AAC
Format/Info : Advanced Audio Codec
Format profile : HE-AAC / LC
Codec ID : 40
Duration : 2 h 48 min
Bit rate mode : Variable
Bit rate : 62.6 kb/s
Maximum bit rate : 77.7 kb/s
Channel(s) : 2 channels
Channel positions : Front: L R
Sampling rate : 48.0 kHz / 24.0 kHz
Frame rate : 23.438 FPS (1024 spf)
Compression mode : Lossy
Stream size : 75.3 MiB (99%)
Writing library : Nero AAC codec 1.5.4.0
Encoded date : UTC 2017-09-22 01:08:31
Tagged date : UTC 2017-09-22 01:11:42

[tebasuna51]

That is a typical problem with captures from TV.

When MediaInfo/eac3to read the first header of AC3 show:

Format : AC-3
Duration : 56 min 46 s
Bit rate : 384 kb/s
Channel(s) : 2 channels
Stream size : 156 MiB

All seems correct and we can expect a decoded wav (64 bits) size of:

3406 sec. x 2 ch. x 8 bytes/ch x 48000 samples/sec = 2495 MB

But the decoded wav is:

Caution: The WAV file is bigger than 4GB.

because only the first's AC3 are 2.0 and the rest are 5.1, then the decoded wav have 3 times (6ch/2ch) size and are buggy (noise) recoded to a m4a with a duration of:

3406 sec x 3 = 2h 50m

near the duration 2 h 48 min of m4a (seems there are a few of AC3 2.0 frames, maybe commercials)

An AC3 with a mix of 2.0 and 5.1 frames are invalid and can't be recoded properly, you need obtain valid AC3 from it.

You can use DelayCut or a tool I wrote: SplitAc3 (https://forum.doom9.org/showthread.php?p=1447695#post1447695)

Other related threads:
https://forum.doom9.org/showthread.php?p=1679327#post1679327
https://forum.doom9.org/showthread.php?p=1768231#post1768231