Since you’re using VLC do this:
Open each audio file, go to Tools > Codec Information or just press Ctrl+J.
Do that for both your source file and your exported file from Shotcut. Take a snapshot of that Codec Information window of both your source file and exported file and post them here.
Those were already posted by Emil; that is what I was responding to.
If you are trying to preserve quality do not use AAC both out of Reaper and Shotcut.
Even at a high bitrate like 512k or 640k?
Putting aside the lossless codecs, what then? AC-3?
What should I use instead?
Reaper audio codec options are quite limited, plus there are no exact matches with Shotcut audio codec options (aside from aac)
I am a bit confused why I should change anything in Reaper, because actually I am fine with the rendered audio. Just want to keep this exact audio track after final video editing in Shotcut.
Basically, your audio is suffering from almost every conversion error possible. The good news is that your ears were sensitive enough to detect it.
The output settings in Reaper need to change because it’s exporting in a lossy compressed format. This format will be re-encoded by Shotcut for the video export, then re-encoded again by YouTube or whatever delivery method is being used. There will be accumulative quality loss at each stage. This transcoding chain is the reason that the exact audio exported from Reaper cannot be preserved all the way to the final viewer. You will have zero control over what YouTube does to your audio when it re-encodes.
So it’s best to over-prepare. Sounding “good” out of Reaper is not good enough to survive two more generations of encoding. It has to sound absolutely perfect out of Reaper to get highest-quality audio.
The workflow generally looks like this:
-
In Reaper, the export settings should be set to whatever the source material sampling rate is. If the audio was recorded at 44.1kHz, then the export should be 44.1kHz. Converting from 44.1kHz to 48kHz or vice versa during export will introduce resampling artefacts.
-
In Reaper, the export settings should be a lossless format like 24-bit PCM or 32-bit FP (floating point). The previous settings of AAC 128kbps are devastatingly too low to survive two generations of transcoding. AAC is very inaccurate to begin with, and going down to 128kbps makes a terrible situation worse.
-
In Shotcut, go to the Export panel > Advanced button > Audio tab and set the sampling rate to the same as the Reaper export. Same situation… if Reaper exported 44.1kHz but Shotcut is encoding at 48kHz, there will be resampling artefacts created.
-
In Shotcut, still on the Audio tab, avoid using AAC if music is critical to you. Use AC-3 at 448kbps if exporting as MP4. If you’re okay with exporting as Matroska (MKV), then audio could be exported as FLAC for lossless audio. It depends on how demanding your quality requirements are.
The reasons, charts, graphs, etc behind these recommendations are detailed in this similar thread:
Crackling/Distorted Audio in Exported Video
There is no reason to go to 96kHz(*). Any quality analog-to-digital converter would have oversampled during recording and made a clean 48kHz or 44.1kHz audio file. Once in the digital domain, additional anti-aliasing (rolloff) filters are not applied so long as the sampling rate doesn’t change. There is no reason for more rolloff because a digital source can’t generate a frequency higher than the destination can capture when the sampling rates are equal. Since the main benefit of 96kHz is shifting the rolloff filters into the inaudible range, the absence of filters likewise means the benefits of 96kHz are greatly reduced.
(*): A studio setting with a pristine setup… maybe 96kHz has a place during processing. But the difference between 48kHz and 96kHz will never be noticed in the average consumer’s home.
3 Likes
I was suspecting that.
That was my reaction as well, when I saw that.
Wow, it will go that high! Great!
I went looking for that but couldn’t find in in any preset.
I stand corrected.
My experience with such things dates back to the 80’s, when we were making pro-audio consoles for Disney and Universal, and the transition from analog to digital was just beginning. Much has been learned since then.
I certainly can’t hear any such difference.
@emil_cress has been blessed with an incredible pair of ears.
Very cool! On the side, I do recording engineering and mixdown for jazz and classical musicians. When it comes to audio, I don’t build the airplane like you do, but I know how to fly it!
I know how to build them, but I don’t know how to fly them.
Theoretically I do, but in reality, I am not very good at it. I have a “tin ear”. I have been sound engineer for live performances for Amy Grant and Barry McGuire, and a few others, but that was only because they didn’t have someone else. LOL.
This was my pride and joy, the Harrison PP1 that did all of Disney’s movies for three decades, and its mate at Universal. Seventy-some-odd microprocessors to control it, and I programmed every one of them. (I think that is the one at Universal; the Disney one had four sections.)
2 Likes
Wow! I’ve actually read articles about the PP1 and the acoustic treatment of that room, so it’s cool to meet some of the brains behind it. Definitely worthy of being a pride and joy.
In order to get the summing-bus noise down from -101 dB to the spec -104 dB, I had to put a different prime number in the software of each input module, so that the power surges on the +5dc supply as the latches for the D-A converters triggered would not synchronize. The randomness brought the digital “chainsaw noise” down another 3 dB. That fix was one of the bits that I was proudest of.
3 Likes
Even if that conversion is being done through a lossless codec like PCM or FLAC?
Yeah, I figured AC-3. Seeing as that AC-3 can go as high as 640kbps, is 448kbps a choice if you want to preserve quality for another generation without getting as big a file size from a 640kbps export?
Is choosing an AAC encode for export going to lose quality even if the bitrate is as high at 512k or 640k?
And in the list of audio codecs at Shotcut’s Export menu is eac3. Is that ffmpeg’s equivalent to Dolby’s E-AC-3 or is it the exact E-AC-3 from Dolby?
1 Like
I know a little bit about some things that should not be done in some audio processes, but I didn’t know exactly why that was the case.
Thanks for this explanation. Now I understand some specifics of AD oversampling in my equipment and also why it is not so interesting to force oversampling in digital.
And it is also important as you say, the final use of that audio.
My intuition is that YT for audio is sufficient for the video consumer type of platform, but definitely not for audio consumers.
1 Like
Correct. The codec is lossless, but the conversion process itself is not. The information that goes in is not the same as the information that comes out. Converting from 48 to 44.1 means information is being discarded, which leads to aliasing errors that are similar to ringing and halos when downscaling in the graphics world. Converting from 44.1 to 48 means information is being created through guesswork, which causes audio smearing that is similar to the blurriness of an image after upscaling it by 2x. Likewise, saving an upscaled image as PNG (lossless) will prevent further deterioration, but the upscale process itself already ruined the image.
Good point… I’m in the habit of saying 448kbps because some hardware devices like Blu-ray players will reject anything over 448kbps, so that’s the recommendation for maximum device compatibility. But if there’s no chance of hardware playback and the goal is to protect the audio against another generation, then yes, 640kbps makes more sense. The size difference between 448k and 640k is peanuts compared to how much space the video codec takes up, assuming somebody wants high-quality video to go with their high-quality audio.
Yes, significantly. The issue isn’t the bitrate. The issue is the quantization and psychoacoustic model used for reducing frequencies and patterns that the human ear can’t detect in order to get higher compression gains. They’re too aggressive compared to other codecs, and the reduction actually can be detected on many types of source material. In the other thread linked above, there are some graphs showing where AAC introduces a wobble into WAV audio files that have nothing in them but a steady sine wave. Increasing bitrate didn’t make the wobble go away. AC-3 and Opus were far more accurate to the source.
Caveat: When we say “AAC” in the forum, we mean “ffmpeg AAC” unless otherwise indicated. Meaning, a different implementation of AAC by a commercial company may not necessarily have the same issue. But it’s still generally accurate to say that AC-3 is less abrasive than AAC when it comes to making reductions for compression gains.
It is an ffmpeg recreation. In the past, it had decoding issues on some hardware devices like TVs. I’m not sure what the current situation is. Standard AC-3 is the better option unless 7.1 audio is needed.
5 Likes
@ Austin
Thanks for the in depth explanation, much appreciated! Definitely will check out your suggestions.
I have already compared (by ear only) audio playback quality from Reaper with later stages after rendering, when reproduced in different media players and uploaded on Youtube. You are absolutely right, audio quality always suffers a bit with each further step, regardless what output device is used.
Just to be clear, I don’t want to achieve the ultimate hi-fi experience and I am far from audiophile snobbery, only want to get the best sound possible (within my modest capabilities)
Guess the easiest solution would be to change sequence of working:
- Video editing Shotcut -> render video
- Import video into Reaper -> audio editing (original audio track has been recorded in Reaper anyway)
- Final rendering out of Reaper
2 Likes
Nah, I just use nearfield monitors with good frequency balance, I am sure you would hear the difference too. I am 50+, wish I had my ears when I was 20 
1 Like
70+
Too much loud factory
Too much quinine
70+, got my first hearing-aids 4 months ago!
1 Like
This is the part that worries me. Does Reaper re-encode the video? Or does it remux the video by removing the previous audio track, inserting your new audio track, and not touching the video at all? If it’s a remux, all is good. But if it’s re-encoding the video, then you’ve only shifted the accumulative loss from the audio world to the video world, and hurting the video will be much more noticeable to the audience. Video also breaks down more per generation than audio does (with lossy codecs, at least).
That’s a very good point though. I didn’t realize that. Thanks for bringing it up.
So is AC-3 just all around superior to AAC? Is the only thing that AAC is good for is that it’s the codec that some devices uses and sites like youtube?
Are these issues you mentioned for AAC even worse on MP3? Is MP3 still preferable over AAC for something?
1 Like