Audio DTS conversion

[eddy89]

First of all, hello everyone, this is my first post. Even though I have been consulting the forum for a while.

Lately, due to my NAS space limitations, I have been encoding/converting a lot of my movies to reduce space:
- Some of them changing from a x264 video to another x265 encode, which most of the times the size is reduced having about the same quality. Even though video is not my concern here, so forget anything regarding video here on this post.
- Some of them removing the DTS/DTS-HD/DTS-MA tracks to a AC3/AAC encode (as I have tracks of 3/4gb that could be converted to 500-600Mb), which is my concern.

Don't need much technical info. I know how to use mkvtoolnix, handbreak, audacity, delaycut, ffmpeg (mostly), etc... so is more a question about what pipeline to follow rather than explaining how to use it (I can learn myself how to use a tool). I just want to know the better way to go.

Now my stupid question, that I am sure it is being asked hundred times (although I was not able to find on the forums):
What could be the better program/technique to encode from DTS/DTS-HD/DTS-MA to AC3/AAC?

Since several days, I have been opening the DTS/DTS-HD/DTS-MA into audacity and then export to AC3/AAC, but I was thinking that maybe this is not the way to go for any given reason.
Is this the better way to go? Is it better to just use ffmpeg or eac3to instead? I am not even sure if audacity uses ffmpeg internally...
Is there any other better way to do it? Am I doing something wrong -or inefficient- using audacity just for conversion?

How I should proceed? Should I change my approach for converting audio?
Wanted to mention, I am talking just using free tools.

Thank you very much in advance guys,

E.

P.S.: sorry for the lame question

[pandy]

I recommend ffmpeg but seem you missed in case of AC3 problem with channel layouts higher than 5.1 - DTS is also max 5.1 but different speaker placement so probably AAC seem to be better choice but still there are some limitations for Encoder libfdk_aac [Fraunhofer FDK AAC]: Supported channel layouts: mono stereo 3.0 4.0 5.0 5.1 6.1(back) 7.1(wide) 7.1 5.1.2(back) .

Btw Audacity use ffmpeg https://support.audacityteam.org/basics/installing-ffmpeg

[eddy89]

I recommend ffmpeg but seem you missed in case of AC3 problem with channel layouts higher than 5.1 - DTS is also max 5.1 but different speaker placement so probably AAC seem to be better choice but still there are some limitations for Encoder libfdk_aac [Fraunhofer FDK AAC]: Supported channel layouts: mono stereo 3.0 4.0 5.0 5.1 6.1(back) 7.1(wide) 7.1 5.1.2(back) .

Btw Audacity use ffmpeg https://support.audacityteam.org/basics/installing-ffmpeg

First of all, thanks for your reply pandy.
I never converted anything higher of 5.1 to AC3, in this case I always use AAC, so there is no problem here.
Not sure if audacity uses ffmpeg for any kind of conversion, because it seems to use it in certain cases, it does not mention that is using it for AC3 for example..., so I am not completely sure about it.

What I am more concern is that if using audacity, is does kind of double conversion:
- When you open it, it looks like is processing it, for converting/decoding the input format to a kind of PCM format (in which you could see the representation of the wavelength there, right?).
- Then having that PCM format in the middle, it just converts to any given format.
Maybe I am assuming too much, and is not what is happening. Please, anyone, correct me if I am wrong.

But if does, I am not sure if this kind of dual conversion is "optimal", or on the other hand is harmless. Which in that case I am fine with it (if the result of converting with audacity has the same quality than using directly ffmpeg or there is a penalty because of it.

Maybe someone can tell me if what I am doing is proper or not, or should I use a more direct method instead (like ffmpeg or ac3to).

I need to know this before keep proceeding in the way I am doing right now, cause I encoded like 20+ movies already (which I removed the sources, so I can't put it back), and I have the fear to lose content if I am not doing well.

Thanks guys in advance again.

[pandy]

Conversion from let say 'DTS' codec (more correctly internal signal representation) to another codec (for example AAC) is usually performed trough intermediate step using PCM - this is accepted as 'optimal' way of conversion (under some constrains).

Audacity is audio editor so unless you don't need additional edition (signal modification) then ffmpeg can offer same or better quality - also as ffmpeg is running in batch then processing can be faster and semi automated.

[eddy89]

Conversion from let say 'DTS' codec (more correctly internal signal representation) to another codec (for example AAC) is usually performed trough intermediate step using PCM - this is accepted as 'optimal' way of conversion (under some constrains).

Audacity is audio editor so unless you don't need additional edition (signal modification) then ffmpeg can offer same or better quality - also as ffmpeg is running in batch then processing can be faster and semi automated.

Thanks pandy, such a great reply!

Now I am kind of understanding some things.

If I understood correctly, what a relieve!! that the audacity process does not make any additional step, and therefore not cause any penalty, quality wise (which means that I haven't lost my sources -deleting after processing them- doing a conversion process wrong).

Assuming that the decodification process perse from any signal representation transforms to an intermediate PCM signal in order to then transform to another signal representation (like AC3), then we could say that the quality of audacity/ffmpeg must be the same -assuming the algorithm is the same-, right?
So please let me know about the better quality you mentioned of the ffmpeg:

can offer same or better quality

I guess I have to assume then, that ffmpeg works with better algorithms which could improve the quality of the output signal?, cause I was assuming the same quality conversion algorithms and not less.

On the other hand, I understand the optimisation part you mentioned, with batch automation and even multiprocessing, which I am sure that having a lot of conversions, would matter a lot. In my case as I am going 1 by 1, I am not very concern about this thing. I was using audacity for the convenience of having a UI and possibly if anytime I have to edit something (like tempo/speed change or other things), which in most of the cases I will not need to.

By the way pandy, do you use ffmpeg directly with command line, or you use any UI with it? I use an intel macOS laptop by the way.

Thanks, I appreciate the information

Edit: I tried the ffmpeg GUI Shutter Encoder, and is very easy, very quick and clean UI to use (maybe a little bit too much just for audio as it does a lot more things and therefore the size is not small). I am open to suggestions though

Edit2: I tried the same conversion with Audacity and Shutter Encoder and produce exactly the same size file (and opening with media info has EXACTLY the same information). Even though not sure if it is exactly the same file produced...

[hello_hello]

Try this program. It's portable so you can just unzip it somewhere and run it.If I remember correctly I put the Encoders folder on the root of the C drive. If you attempt a conversion though and the program doesn't know where to find the relevant encoder, it'll ask you to show it where it is.

https://forum.videohelp.com/threads/396860-foobar2000-portable-%28for-audio-encoding%29

You can load all sorts of files into a playlist. It'll even accept MKVs and MP4s containing video and decode the audio within. You can convert multiple files simultaneously and even convert the same files to different formats at the same time if your hard drive and CPU can keep up. For AAC I'd recommend using the QAAC encoder presets. The one labelled "no padding, single thread" is a preset that enables the QAAC option to remove the padding from the beginning of the encoded file (lossy enocders add a little silence to the beginning and end of the audio). It'll only encode a single file at a time. The preset labelled "no padding" will encode multiple files simultaneously.

I didn't have anything containing DTS audio handy as I always convert it, but this is the idea. Foobar2000 decodes lossy audio as 32 bit float, optionally runs it through a DSP chain if you tell it to, and most of the encoders accept 32 bit float as the input audio. Foobar2000 does it's decoding with ffmpeg, although you'll need to provide it a copy to use the presets that encode with ffmpeg. I didn't include it in the zip file to keep the file size down.


[Attachment 86344 - Click to enlarge]

[hello_hello]

I use an intel macOS laptop by the way.

I didn't read that until after I submitted my previous post. There's is a Mac version of foobar20000 but I don't think it's as fully featured as the Windows version. Not that I run foobar2000 on Windows anyway. I'm running the Windows version in Wine on Linux.

[cholla]

Audacity should automatically locate ffmpeg.
If you want to check if Audacity has done this you can:
Go to Edit > Preferences > Libraries
Click on the Locate... button.
Then select Yes you want to locate manually.
You can use what Audacity finds of Cancel.
If no ffmpeg is found then do this:
In the dialog window, click Browse... to locate the avformat-*.dll from the FFmpeg folder you downloaded/installed elsewhere
Once you've found it, click Open, then OK, then OK again to close the preferences.

I use this
https://lame.buanzo.org/ffmpeg.php
You might check this also:
https://lame.buanzo.org/

This is the ffmpeg message I get when I open an DTS file:

[Attachment 86353 - Click to enlarge]

[pandy]

If I understood correctly, what a relieve!! that the audacity process does not make any additional step, and therefore not cause any penalty, quality wise (which means that I haven't lost my sources -deleting after processing them- doing a conversion process wrong).

Audacity is audio editor and also trying to keep quality highest as possible so in sense of quality there should be no substantial difference in your results.

Assuming that the decodification process perse from any signal representation transforms to an intermediate PCM signal in order to then transform to another signal representation (like AC3), then we could say that the quality of audacity/ffmpeg must be the same -assuming the algorithm is the same-, right?

Yes, but as always there can be subtle differences leading to quality differences (but those quality differences may be not audible so from listener perspective quality is similar).

So please let me know about the better quality you mentioned of the ffmpeg:

can offer same or better quality

I guess I have to assume then, that ffmpeg works with better algorithms which could improve the quality of the output signal?, cause I was assuming the same quality conversion algorithms and not less.

Yes, for example there are multiple encoders for AAC - some of them use different license and can't be distributed in binary form as integral part of ffmpeg.exe (but they can be distributed as independent library). For example considered as high quality AAC encoder 'libfdk_aac' [Fraunhofer FDK AAC]. Normally ffmpeg is equipped with own AAC encoder 'aac' but accordingly to different listening test it is worse than 'libfdk_aac'.
This is where with proper ffmpeg build you may achieve higher quality than Audacity.

On the other hand, I understand the optimisation part you mentioned, with batch automation and even multiprocessing, which I am sure that having a lot of conversions, would matter a lot. In my case as I am going 1 by 1, I am not very concern about this thing. I was using audacity for the convenience of having a UI and possibly if anytime I have to edit something (like tempo/speed change or other things), which in most of the cases I will not need to.

What is more important you can transcode audio track without touching video and at the same time mux them together so except size reduction practically transparent conversion.

By the way pandy, do you use ffmpeg directly with command line, or you use any UI with it? I use an intel macOS laptop by the way.

I use batch - it provide better control especially important for ffmpeg as ffmpeg evolve very quickly and it is difficult for GUI developers to introduce those changes quickly. GUI is fine for typical operation but batch file after some practicing can be faster and more flexible.
As always this is indvidual decision - when i was young preferred GUI's over batch, now when i'm old i prefer batch over GUI.

[eddy89]

Thanks hello_hello and cholla. Unfortunately the replies doesn't apply to macOS. Thanks anyway.

If I understood correctly, what a relieve!! that the audacity process does not make any additional step, and therefore not cause any penalty, quality wise (which means that I haven't lost my sources -deleting after processing them- doing a conversion process wrong).

Audacity is audio editor and also trying to keep quality highest as possible so in sense of quality there should be no substantial difference in your results.

Assuming that the decodification process perse from any signal representation transforms to an intermediate PCM signal in order to then transform to another signal representation (like AC3), then we could say that the quality of audacity/ffmpeg must be the same -assuming the algorithm is the same-, right?

Yes, but as always there can be subtle differences leading to quality differences (but those quality differences may be not audible so from listener perspective quality is similar).

So please let me know about the better quality you mentioned of the ffmpeg:

can offer same or better quality

I guess I have to assume then, that ffmpeg works with better algorithms which could improve the quality of the output signal?, cause I was assuming the same quality conversion algorithms and not less.

Yes, for example there are multiple encoders for AAC - some of them use different license and can't be distributed in binary form as integral part of ffmpeg.exe (but they can be distributed as independent library). For example considered as high quality AAC encoder 'libfdk_aac' [Fraunhofer FDK AAC]. Normally ffmpeg is equipped with own AAC encoder 'aac' but accordingly to different listening test it is worse than 'libfdk_aac'.
This is where with proper ffmpeg build you may achieve higher quality than Audacity.

On the other hand, I understand the optimisation part you mentioned, with batch automation and even multiprocessing, which I am sure that having a lot of conversions, would matter a lot. In my case as I am going 1 by 1, I am not very concern about this thing. I was using audacity for the convenience of having a UI and possibly if anytime I have to edit something (like tempo/speed change or other things), which in most of the cases I will not need to.

What is more important you can transcode audio track without touching video and at the same time mux them together so except size reduction practically transparent conversion.

By the way pandy, do you use ffmpeg directly with command line, or you use any UI with it? I use an intel macOS laptop by the way.

I use batch - it provide better control especially important for ffmpeg as ffmpeg evolve very quickly and it is difficult for GUI developers to introduce those changes quickly. GUI is fine for typical operation but batch file after some practicing can be faster and more flexible.
As always this is indvidual decision - when i was young preferred GUI's over batch, now when i'm old i prefer batch over GUI.

Brilliant reply, pandy!
I think, the apple codec (aac_at) is even better than the Fraunhofer, so I think I will try to use this one and as I am already in macOS I don't need to install anything extra for it.

Do you have any cheatsheet of the batch commands you usually use for converting audio?

For example, now I am trying to convert a DTS-HD MA 7.1 to AAC, what could be the best command line for it? Best going CBR or VBR?

ffmpeg -i /path/audio DTS-HD\ MA\ 7.1\ @\ 2649\ kbps.ENG.dts -c aac_at -aac_at_quality 0 mynewaudio.m4a

I am having problems with the channels, as it is outputting 7 channels instead 8, which is weird.

Thanks