Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: xHE-AAC : The Death of OPUS? (Read 35739 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.


Re: xHE-AAC : The Death of OPUS?

Reply #52
To LithosZA:
"Name any other lossy codec that performs better at those low bitrates."
I was talking about standard AAC, better known as AAC-LC.
I consider low bitrates to be 64 kbit/s and less. AAC-LC is just bad there.
Out-of-box Opus sounds amazing at that bitrate.

"Why does it matter if a lossy codec doesn't have a lossless mode?"
It doesn't matter. But if they want to release real MP3 killer then it needs to have more STANDARD features.
You cannot just keep upgrading existing codec and keep the same name just because of popularity.

To peskypesky:
Again, I was talking about standard AAC-LC. It sucks at low bitrates.
If you cannot hear the difference between 64 kbit/s AAC-LC and lossless then you have some serious hearing problems.

I am not saying that AAC is bad in any way, I just don't agree with their naming.
If device says that it can play MP3 files that means it can play MP3 files.
If device says that it can play AAC files, you have no idea what types of AAC files it can actually play.

JPEG is the most popular image format in the world.
Imagine if they released new version of JPEG but kept the same .jpg extension.
Suddenly you have millions of devices that "support" .jpg but actually they don't.
But they did it properly. They released the updated JPEG-2000 and changed the extension to .jp2.
They avoided confusion.
gold plated toslink fan

Re: xHE-AAC : The Death of OPUS?

Reply #53
They avoided the confusion, and also avoided becoming a successful replacement. Windows doesn't support the format out of box. macOS does, but nobody knows this, so nobody uses it. Linux probably does thanks to many open source projects being bundled with desktop environments, but again, nobody uses this stuff.

WebM and WebP are supported out of box by Windows 10. WebP is supported by macOS Big Sur and iOS 14, as well as third party browsers. Linux, in all likelihood, supports these just fine. They're probably a way better successor to JPEG than anything else right now, especially with their successful market penetration.

HEIC/HEIF is "sort" of a successor format. iOS tried to make it the default, but you had to buy new devices to expect anything to start generating them at all. And everything just converts things to JPEGs when you step outside of macOS anyway. Windows 10 does support these natively, if you install the right stuff from the Microsoft Store, free if your video card supports HEVC decoding, or for a really small fee to license a software decoder. Linux has a lot of trouble with this one, mostly down to either Git/source plugins for GTK/Qt, or GIMP importing from the format.

There are a lot of other random formats that have tried to declare themselves as successors to various things, but none of them really have market penetration or support much of anywhere other than their respective developers' web sites.

Re: xHE-AAC : The Death of OPUS?

Reply #54
jpeg is just like mp3: good enough and supported everywhere. The average user associates music file = mp3, image = jpeg. They open everywhere, there is no need to know more.
As I see it, there will be no new mp3, or jpg equivalent for a long time no matter how much better the new codecs are.

The success stories we see are when enconding, delivery and decoding are done trasnparently to the user. The user desn't have to do or know anything. This is youtube using opus, discord and video call apps using opus for low latency, spotify with vorbis or netflix switching to xHE-AAC.

jpeg has a succesor format that might take over on the long run. It allows lossless repack of jpeg images, has good lossless and lossy compression, and more features. Success, if it eventually happens, will be transparent to the end-user: a website sends the image and the browser displays it... and the user doesn't notice any difference.

JPEG-XL website: https://jpeg.org/jpegxl/

Re: xHE-AAC : The Death of OPUS?

Reply #55
I fully agree with both of you, especially the part about transparent delivery and decoding but this is far from ideal solution.

Quite a lot of people prefer offline storage for various reasons.
Let's start with Spotify, YouTube and other streaming services.

How many times has your playlist been "decimated" by streaming services?
Yesterday you had 100 songs in your playlist, tomorrow you have 90. 10 of them were removed for unknown reason.
This cannot happen with offline storage.
What about TV shows and movies?
Yesterday streaming service A had movie B, tomorrow that movie was bought by streaming service C and there is no way to watch it.
You are forced to buy another useless subscription service.
With offline storage you can watch anything anytime.

Now that kode54 mentioned WebP.
Two years ago, one of the most popular car selling websites in my country used WebP. Now they are back to using JPEG again.
People used to download pictures and share with their friends but after WebP update they couldn't because noone of the applications outside of Chrome recognized WebP.

Arkhh says that music is MP3 and image is JPEG. That is 100% correct. For most people MP4 means movie.
I cannot remember how many times people have asked me why their TV is not playing MP4 files.
Have fun explaining them the difference between H.264/H.265, different profiles (baseline, main, high...) and levels.

I'm just saying that we need more "simplified" codec that simply works so people don't have to spend next five hours searching on the internet why "supported" files are not working.
gold plated toslink fan

Re: xHE-AAC : The Death of OPUS?

Reply #56
For most people MP4 means movie.
I cannot remember how many times people have asked me why their TV is not playing MP4 files.
Have fun explaining them the difference between H.264/H.265, different profiles (baseline, main, high...) and levels.
In my opinion, the problem here is not on the computer science side. While it is true that the vast majority of users are unaware of the difference between a container format and a video coding format, this - despite being unfortunate - does not mean that this approach to storing video/audio itself is wrong. It would make no sense for every video coding format to have its own container format. In fact, in my opinion, we already have too many.

Re: xHE-AAC : The Death of OPUS?

Reply #57
jpeg is just like mp3: good enough and supported everywhere. The average user associates music file = mp3, image = jpeg. They open everywhere, there is no need to know more.
As I see it, there will be no new mp3, or jpg equivalent for a long time no matter how much better the new codecs are.
...

Sadly I think you may be right...  Unfortunately though, whilst people are quite content to download an image in whatever format  (jpeg / png / webp etc) and just look at it,  the situation is often different  for audio. ---  Multiple times a week I see people with perfectly usable AAC audio (because of MPEG, ISO, and Industry standards / popularity)  but they insist they need to convert it to MP3 -- just because that's what they're used to. Often they'll be turning 96 - 128k AAC -LC into a CBR 320 , or V0 MP3 because they (think they) know that's "best" , so lose quality in the transcode, risk more artifacts, and waste a ton of space on their precious  iDevice.

Re: xHE-AAC : The Death of OPUS?

Reply #58
Quite a lot of people prefer offline storage for various reasons.
Let's start with Spotify, YouTube and other streaming services.

How many times has your playlist been "decimated" by streaming services?
Yesterday you had 100 songs in your playlist, tomorrow you have 90. 10 of them were removed for unknown reason.
This cannot happen with offline storage.
What about TV shows and movies?
Yesterday streaming service A had movie B, tomorrow that movie was bought by streaming service C and there is no way to watch it.
You are forced to buy another useless subscription service.
With offline storage you can watch anything anytime.

Exactly. that's why I never bother with streaming in general, at least for anything I care about and will want to listen to or re-watch here and there as the years pass.

it's always best to have a file stored locally (or physical copy of some type) as then you can always listen/watch it anytime you want. plus, not everyone has a fast internet line and in cases like this, all the more reason to have it locally stored as then ones internet line speed does not matter all that much.

so while some casual YouTube stuff is okay for some basic videos here and there, for music/movies I always prefer to have a locally stored copy, especially when it's a movie/song ill want to see/listen to from time-to-time as the years pass.

Multiple times a week I see people with perfectly usable AAC audio but they insist they need to convert it to MP3 just because that's what they're used to. Often they'll be turning 96 - 128k AAC -LC into a CBR 320 , or V0 MP3 because they (think they) know that's "best" , so lose quality in the transcode, risk more artifacts, and waste a ton of space on their precious  iDevice.

It seems anyone with a basic understanding of audio/video compression would now you can't raise quality of something that's already lowered, it just goes downward with each conversion from lossy-to-lossy.

with that said... while not optimal, as you already know, I could possibly see someone going from say a 320kbps MP3, or higher bitrate AAC-LC file (and the like), to a lower bit rate lossy file occasionally. but, as you already know, it makes no sense to go from something already pretty low on bit rate back to a higher rate, especially from say a AAC-LC 96-128kbps up to 320kbps MP3. so it's always higher-to-lower, not lower-to-higher as a general rule.

but with that said... if I have a rough understanding of how many see this stuff, which probably sums up the group of people your talking about, they probably don't care all that much as long as the audio file still sounds 'good enough' to them. which pretty much means as long as the audio is not obviously bad to them and remains close to what you hear on a radio for example. hence, even doing those 96kbps AAC-LC to 320kbps MP3 probably won't matter to them since the quality will not degrade enough for it to matter to them and they think the conversion worked perfectly okay, even though it just wasted some storage space. but even this won't matter to them when they got storage space to burn as it seems the only time this would matter is if they can't fit the songs they want onto a device at which point they might be forced to learn more about it, which probably won't happen nowadays given storage space of music is dirt cheap compared to what it used to be say about 10-15 years ago or so. like back in the old days when 512MB or 2GB cost quite a bit, bit rate of encoded music was much more important. but once 8-16GB or so started to become reasonably priced, short of people who wanted to store boatloads of music on their device, you could pretty much fit ones entire collection on there at a decent enough bit rate.
For music I suggest (using Foobar2000)... MP3 (LAME) @ V5 (130kbps). NOTE: using on AGPTEK-U3 as of Mar 18th 2021. I use 'fatsort' (on Linux) so MP3's are listed in proper order on AGPTEK-U3.

Re: xHE-AAC : The Death of OPUS?

Reply #59
The state of lossy music encoding is disheartening. Breakthroughs for the benefit of humanity rather than for profit, such as Vorbis and Musepack, faded before reaching an apex because the authors changed the priorities for some reason. LAME also stalled. Yet killer samples to illustrate the flaws are still here. Some audiophiles with coding skills who patched psychoacoustic models disappeared. Dear visionaries who are good at math, has your mojo dried up, swallowed up by age, illnesses, streaming services and the petty bourgeois swamp? Or are all your efforts now focused on improving hash functions and quantum-resistant cryptography? It seems that only @bryant is still kicking with the development of lossless WavPack, although we recently discussed how to refine its hybrid mode. And QOA, the only fresh lossy codec in recent times, gains no traction as if the author released it only to prove the idea of doing better than Vorbis, hoping that Public domain license would encourage others to continue the undertaking.
• Join our efforts to make Helix MP3 encoder great again
• Opus complexity & qAAC dependence on Apple is an aberration from Vorbis & Musepack breakthroughs
• Let's pray that D. Bryant improve WavPack hybrid, C. Helmrich update FSLAC, M. van Beurden teach FLAC to handle non-audio data

Re: xHE-AAC : The Death of OPUS?

Reply #60
The state of lossy music encoding is disheartening. Breakthroughs for the benefit of humanity rather than for profit, such as Vorbis and Musepack, faded before reaching an apex because the authors changed the priorities for some reason. LAME improvement also stalled. Yet killer samples to illustrate the flaws are still here.
You seem to be working under the delusion that it is possible to create a codec that is 100% transparent for all input and for which no killer samples exist. You seem to think this holy grail is only impeded by "lazy developers".

To me, that would be the same as saying developers of lossless codecs are lazy because codecs are stuck at no more than 50% compression generally. There is simply no way to get much better. Same goes for lossy codecs, at some point there is only so little to be gained from further development no-one bothers anymore.

If you want 100% transparency all the time, take a lossless codec. If you want lower bitrates, you'll need to accept the reality that there might be audible distortion.
Music: sounds arranged such that they construct feelings.

Re: xHE-AAC : The Death of OPUS?

Reply #61
@ktf, I'm sorry, but you missed the point. I didn't use the word transparency and didn't set the goal of getting closer to any value of the very transparency. I see killer samples as bug reports that should be taken into account during development, that's all. Point is that promising codecs are abandoned halfway through. Musepack and Vorbis are abandoned, the room for improvement is there, but ignored, which gives the false impression they are outdated.
• Join our efforts to make Helix MP3 encoder great again
• Opus complexity & qAAC dependence on Apple is an aberration from Vorbis & Musepack breakthroughs
• Let's pray that D. Bryant improve WavPack hybrid, C. Helmrich update FSLAC, M. van Beurden teach FLAC to handle non-audio data

Re: xHE-AAC : The Death of OPUS?

Reply #62
What room for improvement? Just because you think there is any doesn't mean there is, that is called wishful thinking. Also, you can't just fix one "killer sample" in lossy codecs without it having negative consequences somewhere else.
Music: sounds arranged such that they construct feelings.

Re: xHE-AAC : The Death of OPUS?

Reply #63
It's funny to see this "codec death" comments.  :D
Fanboys' wishful thinking, quite possibly?
Listen to the music, not the media it's on.
União e reconstrução

Re: xHE-AAC : The Death of OPUS?

Reply #64
I work in the realm of embedded development and electronics engineering.
Products and infrastructure systems that use audio in any shape or form, be id devices that play alert sounds, do some sort of voice communication (like VoIP), or even do some sort of content delivery of streaming audio, tend to use Opus and/or Vorbis.
So it's not a question of *if* there's gonna be an adoption of it, the question has been already answered.

A good example I like to bring up to kinda drive this home: remember when you were using a GPS SatNav in your car 10 years ago? Remember how it played turn instructions? Well, those were Opus files stored on the device, even older devices used Vorbis.
Opus is relatively easy to use, there are no hidden costs attached to it, and it has been in use for a very long time, so there's simply no reason to switch over to xHE-AAC or any other codec for that matter. Modern SoCs make using Opus rather simple, there's no real reason to use a specific codec because it's easier on the micro controller, etc.

Content delivery isn't really where I work at, but I've yet to see people veering away from Opus where it's already implemented. One such place is Youtube.
Twitch and many other content delivery platforms use other codecs, but I don't see them changing over to xHE-AAC from what they're already using. Now I don't see them switching over to Opus either, but I can see how this might actually be something they'd consider in the future, simply because Opus has been established for a long time by now. And yes - it's kinda crazy to say it but Opus is kindof an "old codec" by now... Old doesn't mean bad, and tbh, xHE-AAC doesn't bring enough benefits to the table...

There's one thing where I've observed xHE-AAC to be kind of being viable **perhaps**, and that's DAB-T. DAB-T is already locked-in with AAC codecs, iirc they use HE-AAC v2, but that's not the only one they use. The problem with that is, however, that once consumers bought their DAB-T kitchen radios, it'd require at least a firmware upgrade and sometimes a new DSP to decode a new codec, that's why you don't see a lot of them be rolled out each year. I think in Europe (at least the parts I know of) they're currently at "DAB-T2" which basically means they switched over to a newer (or set of those) codecs. They'd need to switch over to yet another version of DAB-T if they'd want to use xHE-AAC. And even though this might not happen for a couple more years, I can see that happen eventually.

In other words: where hardware lock-in is just so strong it takes years and political decisions to change them, that's where I can see xHE-AAC being adopted. Anywhere else, not so much. This isn't something I feel like having to look into a crystal ball for, but rather it's what we've seen in the last ten years, so I think it's sensible to assume those trends will pertain at least another five to ten years.

After that, we might have a bunch more codecs to look at, who knows.

Re: xHE-AAC : The Death of OPUS?

Reply #65
You work as an electronics engineer but you don't seem to know the industry regulations very well. MPEG-D USAC (the real name of xHE-AAC) is already part of some TV and DRM radio standards. When it was proposed, it had on paper everything needed for broadcasting.

In my opinion we must ask ourselves whether we are not also part of the problem; for example, I am an early adopter and I tend to evaluate products, especially open-source ones, at a non-optimal stage, it also happened here with Exhale, which didn't convince me, I published some tables (which ended up in the trash) in which I tried to compress the voices at different sampling frequencies and bitrates to see how the encoder behaved, which I use very little today (I use a paid encoder), because I didn't find a solution for archiving (at high bitrates the advantages disappear compared to LC-AAC and at low bitrates the compatibility for the common user was not satisfactory).

Today, after some time, the situation on the compatibility front has remained approximately the same, Windows 11 has not taken the place of Windows 10, leaving the absolute majority of users unable to use xHE-AAC without installing third-party software. At the same time the diffusion of Linux and various BSDs has doubled, and even in this case third-party software will be needed. It goes better with mobile phones, tablets and smart speakers where the user almost never encounters problems, e.g. Sonos speakers receive the decompressed audio and work around the problem.

We come to Opus which in the meantime is not dead, on the contrary it has constantly improved and in my opinion today its strong point is precisely that it can be used everywhere, what was missing was the iPhone which has a significant market share in smartphones and which now reproduces it both in ISOBMFF than in Webm better than in the more efficient CAF. On the Android front, all the non-user upgradeable models that required an extension other than Opus are ending up in landfill, and therefore the situation has improved.

As you can see there are external factors that are much more important for those who produce content than our mental wanks, and we get to today where the first strong point of Opus is that you can compress and decompress on any device, on the user side MPEG-D has the advantage USAC, production side wins Opus which also works well with the recent ISA RISC-V, and completes the job in less time (which means costs).

For my loved music I have singular theories, I don't convert it, I keep it in the format in which it is provided to me because any manipulation can introduce an error and a loss of information. I'm not keen on comparing the various encoders because the space is worth less than my time to compare them. For me, listening to music is an activity to be carried out without other competing tasks, otherwise it becomes like listening to the radio, a musical background from which to distract yourself.

However, I work with voices, and in the broadcast sector they occupy a fundamental part of the schedules, and I have to get them to future generations, so I even use MPEG-D USAC to preserve them (rather LC-AAC, not HE-AAC which in my country is not is much loved, it's the main problem with DAB, people till date have preferred to use FM), otherwise the benefits costs are eliminated. From today I could even imagine using Opus, which has advantages (many for me) and fewer disadvantages than before (for others).

The war of formats, now the war of containers, does not exist in the broadcast sector, earlier I gave the example of the not very widespread DRM radio, the use of Opus was foreseen from the beginning, but the greatest advantages can be had on the internet and precisely for predominantly verbal products such as podcasts or others that use a scarce resource in nature, the bandwidth.

To put it bluntly, today the problems are different, the enormous waste of energy in data centers, the use of video when not necessary, the use of generative artificial intelligence by the many who lack their own, today in the country that knows it does not being the most advanced on the planet cause blockages to services and electricity interruptions over large areas because it is not sustainable to have data centers that consume as much as small cities with 200,000 inhabitants.

You can bet on it but in the future the choices will be political and the result of new taxes to contain abuse, the model of the internet as a landfill with content paid for by advertising is unsustainable and destined to change.

Consider that those who wastes more than others in the world will be the last to understand it. After all, they are the same people who haven't understood that the great mass of content present today will disappear tomorrow.

Re: xHE-AAC : The Death of OPUS?

Reply #66
By the way, the codecs (AAC-LC and Opus) that giants like YouTube impose hardly change the preferences of pirates. Having started listening to MP3 on a Rio player back in the 1990s, I still see the dominance of this lossy format on the well-known, little-known and private sharing sites. The only thing that has changed is that FLAC is posted next to it. Curiously, I got a Cowon player in the mid 2000s with Vorbis support out of the box, but this format remained more of a home one due to suboptimal seeking and higher battery consumption; it came in handy for the mentioned lossless.
• Join our efforts to make Helix MP3 encoder great again
• Opus complexity & qAAC dependence on Apple is an aberration from Vorbis & Musepack breakthroughs
• Let's pray that D. Bryant improve WavPack hybrid, C. Helmrich update FSLAC, M. van Beurden teach FLAC to handle non-audio data

Re: xHE-AAC : The Death of OPUS?

Reply #67
Truth is, just because most us here on HA, at a personal level, are a bunch of format skippers (myself included), we cannot automatically naively assume it is that easy (better: tempting) for large corporations and Big Tech names consortia to do so with the same easiness and lightheartedness.

Edit: I daresay they're not batting eyelids at whichever new-kid-on-the-block CODEC pops up now and then as often as we do.
Listen to the music, not the media it's on.
União e reconstrução


Re: xHE-AAC : The Death of OPUS?

Reply #69
I agree. That reply was in the context of using lossy codecs at high bitrates to encode music to achieve transparency for all possible inputs. It seems there is little room for improvement there. At the bottom end, there are still lots of opportunities.
Music: sounds arranged such that they construct feelings.

Re: xHE-AAC : The Death of OPUS?

Reply #70
The state of lossy music encoding is disheartening. Breakthroughs for the benefit of humanity rather than for profit, such as Vorbis and Musepack, faded before reaching an apex because the authors changed the priorities for some reason. LAME also stalled. Yet killer samples to illustrate the flaws are still here. Some audiophiles with coding skills who patched psychoacoustic models disappeared. Dear visionaries who are good at math, has your mojo dried up, swallowed up by age, illnesses, streaming services and the petty bourgeois swamp? Or are all your efforts now focused on improving hash functions and quantum-resistant cryptography? It seems that only @bryant is still kicking with the development of lossless WavPack, although we recently discussed how to refine its hybrid mode. And QOA, the only fresh lossy codec in recent times, gains no traction as if the author released it only to prove the idea of doing better than Vorbis, hoping that Public domain license would encourage others to continue the undertaking.
Developers are not robots or isolated nerds/geeks.
They have lives and families. Life (and regrettably death) happens to them.
These codecs are their creation and the fruit of their efforts. If they can no longer sustain the efforts or are no longer around on earth, the right thing to do with one's hyperpassion for codecs is to pick up the skillsets where they left off and continue investing the same efforts they made. Not make a point of scolding them at every forum opportunity.

The active zone is now in the sub-30 kbps voice region. I don't see Opus going away any time soon.

Re: xHE-AAC : The Death of OPUS?

Reply #71
> If you want 100% transparency all the time, take a lossless codec

In principle nothing stops something like LossyFLAC or maybe even ("up-tuned") Musepack from guaranteeing 100% transparency due to the way they work. (you can basically require them to maintain a certain SNR at all times, no matter how many bits that would take)

Pure lossless-compressed PCM is always going to use more bits than necessary for louder intervals and that's the lowest hanging fruit to shave off some % of bits.
a fan of AutoEq + Meier Crossfeed

Re: xHE-AAC : The Death of OPUS?

Reply #72
The state of lossy music encoding is disheartening.
Developers are not robots or isolated nerds/geeks. If they can no longer sustain the efforts … pick up the skillsets where they left off and continue investing the same efforts they made. Not make a point of scolding them at every forum opportunity.

The absence of a loud and clear message expressing dissatisfaction with the current state of affairs is often mistaken for satisfaction. "People don't complain, so why bother?" And more often than not, I've seen development resume not because someone decided to dust off abandonware out of boredom, but because they faced the pain and showed empathy for those who reported how this or that change would make life easier or happier. Certainly, there are various ways to report: a hit between the eyes and a sugar-coated plea to name a few. Just as there are original editions of Mark Twain and Agatha Christie novels with words of their era like the N-word and orientals, and there are expurgated editions for a fragile, resentful mindset. For example, a few weeks ago I complained that Helix MP3 encoder, officially abandoned 20 years ago, was not capable of handling more than 16-bit input, and the miracle happened. God bless @Case for that.

Mod edit: Defused potentially offensive word.
• Join our efforts to make Helix MP3 encoder great again
• Opus complexity & qAAC dependence on Apple is an aberration from Vorbis & Musepack breakthroughs
• Let's pray that D. Bryant improve WavPack hybrid, C. Helmrich update FSLAC, M. van Beurden teach FLAC to handle non-audio data

Re: xHE-AAC : The Death of OPUS?

Reply #73
The state of lossy music encoding is disheartening.
Developers are not robots or isolated nerds/geeks. If they can no longer sustain the efforts … pick up the skillsets where they left off and continue investing the same efforts they made. Not make a point of scolding them at every forum opportunity.

The absence of a loud and clear message expressing dissatisfaction with the current state of affairs is often mistaken for satisfaction. "People don't complain, so why bother?" And more often than not, I've seen development resume not because someone decided to dust off abandonware out of boredom, but because they faced the pain and showed empathy for those who reported how this or that change would make life easier or happier. Certainly, there are various ways to report: a hit between the eyes and a sugar-coated plea to name a few. Just as there are original editions of Mark Twain and Agatha Christie novels with words of their era like the N-word and orientals, and there are expurgated editions for a fragile, resentful mindset. For example, a few weeks ago I complained that Helix MP3 encoder, officially abandoned 20 years ago, was not capable of handling more than 16-bit input, and the miracle happened. God bless @Case for that.

Mod edit: Defused potentially offensive word.


Its a catch 22 situation. With Mp3 / AAC it can work due to high user demand.  With less popular codecs like mpc, WV, vorbis
the opposite can happen and already small userbase is further decimated . That may slow down development even more.

Re: xHE-AAC : The Death of OPUS?

Reply #74
> If you want 100% transparency all the time, take a lossless codec

In principle nothing stops something like LossyFLAC or maybe even ("up-tuned") Musepack from guaranteeing 100% transparency due to the way they work. (you can basically require them to maintain a certain SNR at all times, no matter how many bits that would take)

Pure lossless-compressed PCM is always going to use more bits than necessary for louder intervals and that's the lowest hanging fruit to shave off some % of bits.

In addtion to lossyflac, loaded mpc  ,High bitrate WV lossy can be used confidently from around 500kbps . something like -b500x4s1 or more.