Using two DTS-HD audio tracks as sources (by importing) - roughly 1.5GB and 1.6GB respectively - I created a custom track for a movie that was the same length and same number of tracks/channels as each of the originals but contained a little extra data - I added small fades, envelopes and tempo changes in 2-3 places, and also removed a little audio. I also mixed each channel inside the two source tracks together with its corresponding track (track 1 to track 1, track 2 to track 2, etc.) but in the vast majority of cases there was no overlap between the two, and there were only a few minutes total of audio that was actually physically mixed together.
However, despite the seemingly small number of changes I made to the data, exporting the finished product using the external command option and the command
ffmpeg -i - -c:a copy "%f"
(i.e. passing through the audio codec), the resulting audio file is a massive 3.3GB. More importantly, I can’t play the exported file at all using MPC-HC when it’s in the .DTS format, despite the fact that I can play both .DTS source tracks fine with MPC-HC.
Does anyone know what’s going on here and how I can fix it?
If the originals were both “regular” DTS, the first thing Audacity would have done is uncompress the work up to uncompressed “WAV” format 32-floating 44100. So the original shows were very probably much larger than you think while inside Audacity. The Export DTS is then the difference between the Audacity show and the Export settings, not the difference between the two original DTS’s and the Export.
The MP3 people run into this. "How come my show is huge compared to the original files?
There is a further problem. If you do manage to get similar filesizes through export management, it’s very probable that the compression sound damage will increase.
I just tried this - though I didn’t need to use FFmpeg because I wasn’t using it for any extra encoding - and it seems that when exporting the .aup project file as 16-bit PCM via WAV, I end up with a file that’s just 44 bytes larger than the 3.2GB file I got from using “external program” with…
ffmpeg c:a copy
Even better, the resulting file is actually playable and the right length, presumably because the container is .WAV this time rather than the .DTS I had used with the first 3.2GB file.
This makes a little more sense given the above reply to Steve, and that some further research shows that the header file for WAV PCM files is, yep, 44 bytes… but this gives me another question of why the passthrough would result in a 16-bit WAV PCM codec rather than 32-bit float PCM, given that each track within the .aup project shows as “Mono, 48000Hz, 32-bit float”, and given that you yourself say Audacity should uncompress to 32-bit floating WAV/PCM.
Additionally, some research seems to indicate that DTS-HD is actually a superior format to WAV PCM 16-bit. Is this true, and if so, is there any particular reason why Audacity converts my original DTS-HD tracks to an inferior format for editing rather than keeping them as is?
Superior in quality. This was the impression I got from Kozikowski’s last reply, which is why I then asked for clarification, but some further research of my own yesterday indicated that it isn’t actually superior, and that DTS-HD MA is simply (proprietary) PCM that’s been losslessly compressed, which Audacity decompresses back to PCM to work on, and that because both are lossless there’s no difference between the quality of the two.
The passthrough outlined in my original post and throughout my replies in this thread - using the “external program” option in the Export dialogue with the command ffmpeg -c:a copy - this copies the codec through to the output file, known as pass-through, which in the case of the PCM that Audacity was using resulted in a raw PCM output that was both very large and refused to play, and this seems to be the answer to the question I asked here.
However, what I’m curious about is, if it is indeed 16-bit PCM that Audacity is working on in this project - which is confirmed by comparing the filesizes of a standard .WAV PCM export and the ffmpeg -c:a copy - why each channel/track’s Track Control Panel inside Audacity says “32-bit float”, which in the standard Export dialogue is listed as being a separate type of PCM to 16-bit PCM. This is now the only thing that I’m confused about, as I now have an answer to all my other questions in this post.
why each channel/track’s Track Control Panel inside Audacity says “32-bit float”, which in the Export dialogue is listed as being a different type of PCM to 16-bit PCM
There are technical/mathematical/programming advantages of using floating-point for digital signal processing. By default, Audacity (like most audio editors/DAWs) uses 32-bit floating-point “internally”, no matter what format you open.
…When you export you have lots of choices.
Additionally, some research seems to indicate that DTS-HD is actually a superior format to WAV PCM 16-bit.
It’s (usually) 24-bits so it’s higher resolution, and it’s (losslessly) compressed so it takes-up less space than WAV. Plus, it’s usually 5.1 or 7.1 channel surround sound from a Blu-ray which “sounds better” than a regular-old stereo CD or WAV file…
But it’s not necessarily audibly “superior” because the guys who do [u]Secintific, blind, ABX tests[/u] have demonstrated that you can’t hear a difference between a high-resolution original and a “CD quality” copy. Or, you could say that CD quality (16-bit/44.1kHz) is better than human hearing.
I didn’t get a notification for this because I wasn’t properly quoted, but thanks for this clarification, that now makes a lot of sense.
Additionally, some research seems to indicate that DTS-HD is actually a superior format to WAV PCM 16-bit.
It’s (usually) 24-bits so it’s higher resolution, and it’s (losslessly) compressed so it takes-up less space than WAV. Plus, it’s usually 5.1 or 7.1 channel surround sound from a Blu-ray which “sounds better” than a regular-old stereo CD or WAV file…
But it’s not necessarily audibly “superior” because the guys who do > [u]Secintific, blind, ABX tests[/u] > have demonstrated that you can’t hear a difference between a high-resolution original and a “CD quality” copy. Or, you could say that CD quality (16-bit/44.1kHz) is better than human hearing.
Good point about the resolution of the file, always worth nothing that that’s what actually dictates the quality of the file, lossless or not. Thanks for the link to the ABX tests, they’re very interesting. I never have expected to hear the difference between lossless copies, the only reason I would ever convert to lossless is to retain the original size of current lossless codecs while compressing. Thanks again.