The sound is different whether I render a wav or mp4

[Ron]

I want to render my videos so that they can be used as universally as possible.
Once on my video channel at tube.tchcns.

Peertube (or at least any sanely admin’ed instance of it), like all the other *tube and streaming services, transcodes uploaded video before serving it to users - so it doesn’t matter what you chose from that perspective, nobody is going to see the file that you uploaded, they’re going to negotiate a format and resolution (from the set the peertube admin selected) that the user’s client likes best.

Don’t you think I’m not on the safest side with aac in mp4 video?

Some Peertube admins prefer unencumbered codecs (in line with the goals of the peertube project itself) - and some pretend codec patents don’t exist, or somehow don’t/won’t apply to them and keep alive the lie that it’s unsafe to use things that the old-world codec mafia don’t control.

But the simple truth is, there is no one-size-fits-all codec choice for A/V use yet.

It all depends on what you want to optimise for. If you can’t strap a modern set-top box to your old TV (or upgrade its firmware) to enhance its capabilities, then maybe that will be what limits you.

If uploading to a streaming service which transcodes is your main goal, I’d be looking at rendering your “master” to whatever format gives the highest quality for the smallest file size - that optimises your upload time, and gives you the best starting point to take the hit of transcoding losses from.

If you do that, you can always still transcode it to a lesser format your TV supports for use with it, without forever penalising all your future uses by rendering your master in the lowest common denominator format.

don’t you want to tell us where your vast knowledge

Appeal to authority is one of the classic logical fallacies used to prop up a weak claim to truth

I’m not important, only sharing and improving best practice is important.

I’m always learning, nothing I have to say is or should be above being questioned or corrected.

[micha]

Hello Ron,
Thank you again for these detailed descriptions.
Very interesting what you said about the streaming services.
We have written a lot about codecs here. But the much bigger difference is the sound editing with the audio effects in Kdenlive.

Are you familiar with this? Unfortunately, the audio tools are treated very neglected here. The best I can find are the ones from Steve Harris’ SWH plugins.

Do you know any better ones, especially good compressors and equalizers?

[micha]

The problem is probably caused by the aac codec. Interestingly, the bit rate seems to have a major impact: 196k is much better than 256k when it comes to low-frequency noise.

aac 256k

aac 196k

and preferably with wav (only audio, no video no aac):

I would be very grateful if someone who is knowledgeable in this area could tell me whether these very low frequencies below -80 dB are actually a problem or just theoretical. Because I simply cannot eliminate them with Kdenlive without also reducing other important frequencies.

[Ron]

@micha

tell me whether these very low frequencies below -80 dB are actually a problem or just theoretical.

Didn’t we already go over all of this?

You’re compressing lossless (wav) audio to a lossy format (AAC) that uses psycho-acoustic trickery to try and make what it lost imperceptible to you. Then working yourself into a panic because math and your eyes can see what the science alleges you can’t hear.

It’s lossy, so by definition it will not be as ‘perfect’ as the original or a lossless copy, and the fewer bit you throw at it, the more lossy it will be - though with a good codec like Opus you can throw away a surprising amount and have it still be impossible for people with Golden Ears to hear the difference. That’s the bitrate at which it is Transparent (and for Opus it’s much less than the rates you’re using here for pretty much any 48k stereo content).

The only way to know if some codec is transparent to you at a particular bitrate on some particular content, with the equipment you are listening to it on, is to listen to it.

If you want to know if it’s transparent, grab yourself one of the tools for ABX testing and see if you can reliably pick the difference between them. But realistically - unless you’re making content to be heard on high-end equipment in an utterly silent sound lounge - just listen to it - if you can hear a problem, there’s a problem. If you can’t, it’s fine, QED.

Normally I’d say anyone who claims to be able to tell you that from looking at the kind of graphs you posted is selling you a bridge. And asking them to tell you from that is begging to be sold one. But in this particular case, your graph looks like ~-46dB white noise - so I’m going to go out on a limb and bet it sounds “Quietly terrible, regardless of what you encode it with”.

If you want a meaningful test, encode some real content, and listen to it with your real ears.

[micha]

Hello @Ron,

The problem reappears when I look at the frequency analysis to see how well my Glame Highpass filter is working: To my horror, I still see this low-frequency data garbage. Despite all my efforts with highpass or asubcut, I couldn’t get rid of it. I was pretty frustrated until I realized that it was due to aac and not Kdenlive: everything was clean with wav or flac.

Of course, I don’t hear these unnecessary frequencies—that’s not the problem. It’s about the unnecessary data garbage that I would like to get rid of.

I am very happy with how the audio in my videos sounds to my ears; that’s not the problem. I just don’t understand where these low frequencies are coming from, which should actually be cut out and are not present in wav either. Do you know if it could be that it’s not Kdenlive, but simply the aac codec?

At the time, we talked a lot about other codecs, especially opus, but what use is a codec that isn’t universally understood? My computer can’t easily play videos with opus audio.

I assume that aac is the safest and most common codec for MP4 videos.

[micha]

Hello @Ron,

You know so much about audio, so I have a big request for you:

1. Is this low-frequency data garbage a problem that I should solve? Does it interfere with playback in any way?
2. Would it be better to reduce the bitrate to 192k with the aac codec, because there is less of this noise, or should I stick with 256k, because that generally offers higher quality?

[Ron]

@micha

what use is a codec that isn’t universally understood?

I have lots of devices that can’t process AAC, does that make it useless too?

My computer can’t easily play videos with opus audio

Citation needed. What doesn’t work with it in 2025?

Is this low-frequency data garbage a problem that I should solve? Does it interfere with playback in any way?

I don’t know what you’re asking that hasn’t been gone over repeatedly.
This is audio - only your ears can answer that.

Would it be better to reduce the bitrate to 192k with the aac codec, because there is less of this noise

Would ‘less’ of something that you say you cannot perceive be ‘better’?
Is that what you’re asking?

How would you ever be able to know? By what measure does something imperceptible become Better or Less? What evidence do you have that it’s ‘garbage’ and not beneficial?

should I stick with 256k, because that generally offers higher quality?

Does it? Can you hear a difference in quality?

Voodoo numbers and internet opinions can’t help you answer this. You have to use your ears. What you hear is the only quality metric that matters.

[micha]

Hello @Ron,

The people who ask questions here don’t have nearly as complex and in-depth knowledge as you do.
I don’t want to question your statements, but sometimes I would prefer a simpler answer.
Presumably, I need to ask clearer questions. So I’ll try again:

After editing in Kdenlive, my video sounds very good to my ears. Probably for everyone else too.

But when I did a frequency analysis to check, I noticed these low frequencies below 50 Hz, which are completely unnecessary for speech transmission. Surprisingly, they are caused by aac, a codec that is very widely used for MP4 videos and is also highly recommended as standard because it makes the video compatible with all kinds of platforms. But since wav doesn’t produce this data garbage, I’m very surprised.

So my question is basically this: Is the technology, the media player, the amplifier, or the speakers in any way negatively affected by the low-frequency garbage? My ears don’t mind it.

[Ron]

@micha

After editing in Kdenlive, my video sounds very good to my ears. Probably for everyone else too.

Then it’s fine. QED. It really doesn’t get any simpler than that.
We told you that a year ago, and nothing about ears or DSP has changed since then.

I noticed these low frequencies below 50 Hz, which are completely unnecessary for speech transmission.

There are lots of things you can’t perceive. That doesn’t make them unnecessary.
You’re only confusing yourself by creating your own arbitrary definitions of ‘garbage’ and ‘quality’ without trusting the only test that matters, your ears.

You’re already throwing way more bits at the problem than you’d need to with a good codec, so:

• If you care about compression, use a better codec.
• If you care about quality - put the analysers that you don’t understand away and listen to what you created.

It really is That Simple.