Audio file format and compression discussion

YouTube · Unread post by **K4sum1** » 16 May 2025, 20:14

Edit: So this kept going, so it gets it's own topic now

This entire post is pretty off topic, but idk where else to post it so.

xperceniol_sal wrote: ↑16 May 2025, 15:19 I keep my XP very VERY slim but I also don't do much and some games lose me because my attention span sucks nowadays. Some days all I can play is my lame racing car games.

We haven't chatted in a while - how are you doing? I sort of miss you but I realize you are VERY busy. I just don't give up on my friends easily and you are one of them

I like to keep my installs slim, but then I get lazy and eventually reinstall anyways so idk. I don't run Steam at startup so it might as well not be there for like 95+% of my computer usage.

I'm doing good, I'm just quite busy, and don't really look here outside of skimming new post notifications. I also find that I keep myself too busy to really think, anywhere from working on r3dfox and such to gaming. Yesterday and today I've been doing a bunch of research into audio formats and stuff, more specifically Ogg Vorbis.

If you're curious:
https://forums.rockbox.org/index.php/topic,54252.0.html
https://github.com/OptiVorbis/OptiVorbis/issues/134

I'm also trying to figure out what the transparency threshold is for Vobis, and if aoTuV is still better than libvorbis today. All the information about this is nearly 20 years old. libvorbis gives me better looking spectrals than aoTuV, but idk if that is necessarily better, and outside of ABX idk how to determine that. However I can't tell the difference between a 128kbps MP3 and a FLAC, so I don't have any hope of being able to ABX this myself so I have no idea what to do. Will probably be another thing I abandon after a day or so idk.

OCCiera · Unread post by **OCCiera** » 16 May 2025, 20:47

K4sum1 wrote: ↑16 May 2025, 20:14 This entire post is pretty off topic, but idk where else to post it so.

xperceniol_sal wrote: ↑16 May 2025, 15:19 I keep my XP very VERY slim but I also don't do much and some games lose me because my attention span sucks nowadays. Some days all I can play is my lame racing car games.

We haven't chatted in a while - how are you doing? I sort of miss you but I realize you are VERY busy. I just don't give up on my friends easily and you are one of them
I like to keep my installs slim, but then I get lazy and eventually reinstall anyways so idk. I don't run Steam at startup so it might as well not be there for like 95+% of my computer usage.

I'm doing good, I'm just quite busy, and don't really look here outside of skimming new post notifications. I also find that I keep myself too busy to really think, anywhere from working on r3dfox and such to gaming. Yesterday and today I've been doing a bunch of research into audio formats and stuff, more specifically Ogg Vorbis.

If you're curious:
https://forums.rockbox.org/index.php/topic,54252.0.html
https://github.com/OptiVorbis/OptiVorbis/issues/134

I'm also trying to figure out what the transparency threshold is for Vobis, and if aoTuV is still better than libvorbis today. All the information about this is nearly 20 years old. libvorbis gives me better looking spectrals than aoTuV, but idk if that is necessarily better, and outside of ABX idk how to determine that. However I can't tell the difference between a 128kbps MP3 and a FLAC, so I don't have any hope of being able to ABX this myself so I have no idea what to do. Will probably be another thing I abandon after a day or so idk.

Why not just use lossless files and a battery bank? I use an Axim x51V as a music player, I keep all my music in WMA lossless and just bring some spare batteries for it, since I have the luxury of a battery hatch.

YouTube · Unread post by **K4sum1** » 16 May 2025, 21:01

Oh, my collection is mostly lossless FLAC with MP3 for the songs that I can't find any lossless version of. I want to have a Vorbis collection for more space constrained devices and reduce transfer times over slow USB 2.0 connection.

The tests are because I'm curious and also it's cool idk. I've found the battery is good enough for even my longest trips with my current collection. I bring a battery bank if I feel I need one, but it's nice to not have to use it. It seems like with this usage perhaps storage uses more power, so maybe a Vorbis collection is best all around for this. I'm also not going to be able to notice the difference especially if I use a phone with Bluetooth, so yeah, but I want to aim for general transparency anyways as it seems achievable. Be it 128k/q4, 160k/q5 or 192k/q6.

Unread post by **Duke** » 17 May 2025, 13:31

OCCiera wrote: ↑16 May 2025, 20:47 I keep all my music in WMA lossless

Why WMAL ? You'd better use FLAC for best compatibilty between various devices

K4sum1 wrote: ↑16 May 2025, 21:01 I want to have a Vorbis collection for more space constrained devices
It seems like with this usage perhaps storage uses more power, so maybe a Vorbis collection is best all around for this.
I want to aim for general transparency anyways
Be it 128k/q4, 160k/q5 or 192k/q6.

I also use Vorbis q5, best quality to my ears with the best compatibilty. MP3 needs 192 kbps to achieve the same quality but it can be played on any device made in the past 25 years, if not more.
Information and discussion about battery usage between various formats:
https://hydrogenaudio.org/index.php/topic,124857.0.html

YouTube · Unread post by **K4sum1** » 17 May 2025, 14:11

So I split this into it's own topic now

Duke wrote: ↑17 May 2025, 13:31
K4sum1 wrote: ↑16 May 2025, 21:01 I want to have a Vorbis collection for more space constrained devices
It seems like with this usage perhaps storage uses more power, so maybe a Vorbis collection is best all around for this.
I want to aim for general transparency anyways
Be it 128k/q4, 160k/q5 or 192k/q6.
I also use Vorbis q5, best quality to my ears with the best compatibilty. MP3 needs 192 kbps to achieve the same quality but it can be played on any device made in the past 25 years, if not more.
Information and discussion about battery usage between various formats:
https://hydrogenaudio.org/index.php/topic,124857.0.html

Yeah. Are you using aoTuV or libvorbis for these files? Also I wonder if you'd be willing to test them if you haven't tried yet. I'd be curious to see if aoTuV is still better.

For the battery usage, I was referring to the Rockbox forum link in the first post, specifically this.

speachy wrote: Yes, 128K vorbis is more expensive to decode than 320K mp3, but that 320K mp3 will wake up the storage device nearly 3x as often. With SSD swaps that's less expensive than before, but even when only reading a uSD card will be in the 100-200mW range. By comparison, one of the ipod5g's CPU cores running at max speed only uses about 20mW in comparison.

You can also see the power consumption tests I ran here.
https://forums.rockbox.org/index.php/topic,54252.msg256140.html#msg256140
https://forums.rockbox.org/index.php/topic,54252.msg256161.html#msg256161

YouTube · Unread post by **K4sum1** » 17 May 2025, 15:50

I did some more research. I found a game that uses 128k Vorbis and the spectrals look really good. However the file seems to have been made by "Sony Ogg Vorbis 1.0 Final" which I would assume to mean some sort of audio software, maybe Sound Forge idk. Could be worth looking into. I also noticed the file was 128 ABR instead of the VBR I've been looking at here. I can confirm that a 128 ABR Vorbis looks better in a spectral, but it is also a bit bigger so yeah. However I noticed aoTuV ABR is functionally identical to aoTuV VBR, at least for 128k. I got two equally sized files with spectrals that match, although the hash is different.

I looked a bit closer and the spectrals for libvobris look a bit harsher or intense in the bits with more sound than the original or aoTuV. I assume this could be noise or artifacts and perhaps why aoTuV is considered to be better when the more extreme shelf looks worse at first when looking at a spectral.

Unread post by **Duke** » 17 May 2025, 22:11

K4sum1 wrote: ↑17 May 2025, 14:11 For the battery usage, I was referring to the Rockbox forum link in the first post, specifically this.

Sorry, I missed the Rockbox bit.

K4sum1 wrote: ↑17 May 2025, 14:11 Are you using aoTuV or libvorbis for these files?
I'd be curious to see if aoTuV is still better.

I'm not using aoTuV anymore because it's supposed to be optimized for low bitrates below -q5 and as I said I'm using q5.
For sure files are smaller with low bitrates but storage is cheap nowadays so I don't care about file size and I prefer quality.

K4sum1 wrote: ↑17 May 2025, 15:50 However I noticed aoTuV ABR is functionally identical to aoTuV VBR, at least for 128k. I got two equally sized files with spectrals that match, although the hash is different.

See above, aoTuV is for low bitrates. So at q5 or above there shouldn't be any (big) difference.

K4sum1 wrote: ↑17 May 2025, 15:50 a 128 ABR Vorbis looks better in a spectral, but it is also a bit bigger so yeah.

K4sum1 wrote: ↑17 May 2025, 15:50 I looked a bit closer and the spectrals for libvobris look a bit harsher or intense in the bits with more sound than the original or aoTuV.

Don't look at audio but listen to it

YouTube · Unread post by **K4sum1** » 18 May 2025, 14:50

Duke wrote: ↑17 May 2025, 22:11 I'm not using aoTuV anymore because it's supposed to be optimized for low bitrates below -q5 and as I said I'm using q5.
For sure files are smaller with low bitrates but storage is cheap nowadays so I don't care about file size and I prefer quality.

Duke wrote: ↑17 May 2025, 22:11 See above, aoTuV is for low bitrates. So at q5 or above there shouldn't be any (big) difference.

Well yeah, but my idea was for you to test again if you could. Even if on paper aoTuV is supposed to be for below -q5, it would be interesting to have it tested.

Also for my case I want the Ogg version for devices with less storage and listening situations where I am less likely to hear any difference. So I'd like to know better where the transparency threshold is, and not have to encode everything again if I get this wrong. (I want to lean lower but be cautious which makes me want to lean more towards q4/128-q5/160 but idk what exactly to do yet)

Duke wrote: ↑17 May 2025, 22:11 Don't look at audio but listen to it

Ok, but the problem is last time I tried, I was unable to tell the difference between a 128k MP3 and FLAC on afaik what is the "best" music device I have with afaik what is the "best" headphones I had. So maybe I shouldn't worry about this and just go q4, but idk I just want to know with modern versions of the encoders.

xperceniol_sal · Unread post by **xperceniol_sal** » 18 May 2025, 15:16

K4sum1 wrote: ↑18 May 2025, 14:50...what is the "best" headphones I had. So maybe I shouldn't worry about this and just go q4, but idk I just want to know with modern versions of the encoders...

Yeah, LOVE my Emerson headphones and it (they) are so clear and crisp at the same time and they were a gift so I don't know what they paid but they usually only run about $40 (I think). Also; I have a decent set of JLab Studio Wired On-Ear Headphones.

However the JLab headphones are very sharp and deep bass and sometimes I (actually) DO want some mid-range.

Get rid of any el cheapo headphones because we appreciate quality music too much; and I need it for my meditation each day.

YouTube · Unread post by **K4sum1** » 18 May 2025, 15:24

Yeah, I haven't used any "bad" headphones for probably over a decade now. When I said "best" there, I meant conventionally considered good by others online. (AKG K371BT (wired because I know BT can mess with quality) is what I used to be exact) I like my current headphones (E600Pro) more than any other even though they're the cheapest ones I've bought and I'm not sure any audiophile would call them good. I used to be in audiophile spaces, and I kinda went against their advice going for this and I'm glad I did.

Unread post by **Duke** » 18 May 2025, 18:49

K4sum1 wrote: ↑18 May 2025, 14:50 my idea was for you to test again if you could. Even if on paper aoTuV is supposed to be for below -q5, it would be interesting to have it tested.

What do you want me to test ? The quality of your encodings ?
If so then I need you to give me the source file and the encoded file so I can compare.

K4sum1 wrote: ↑18 May 2025, 14:50 me want to lean more towards q4/128-q5/160 but idk what exactly to do yet
maybe I shouldn't worry about this and just go q4, but idk I just want to know with modern versions of the encoders.

q4 should be good enough for most people, especially if you listen to music in your car or with small desktop speakers.
But q5 is better! Read more transparent. Or use q4.5

YouTube · Unread post by **K4sum1** » 18 May 2025, 20:24

Duke wrote: ↑18 May 2025, 18:49 What do you want me to test ? The quality of your encodings ?
If so then I need you to give me the source file and the encoded file so I can compare.

I more so meant for you to do it yourself, with music you're more familiar with. If you want me to do it, or me to convert the files for you, I could if you want.

xperceniol_sal · Unread post by **xperceniol_sal** » 18 May 2025, 21:18

K4sum1 wrote: ↑18 May 2025, 15:24 Yeah, I haven't used any "bad" headphones for probably over a decade now. When I said "best" there, I meant conventionally considered good by others online. (AKG K371BT (wired because I know BT can mess with quality) is what I used to be exact) I like my current headphones (E600Pro) more than any other even though they're the cheapest ones I've bought and I'm not sure any audiophile would call them good. I used to be in audiophile spaces, and I kinda went against their advice going for this and I'm glad I did.

Bwahahaha ....

Your low end headphones are my high end. Now I tell ya, I do have super cheap headphones somewhere in the closet (several actually) that cost around $8 to 10 bucks LMAO

Wonderful if all you want is mid and treble, but light on the ears, I guess.

I have an AKG microphone (boom) for drums even though I no longer play the drums - I keep it for keepsake as it was also a gift.

OCCiera · Unread post by **OCCiera** » 18 May 2025, 23:06

Duke wrote: ↑17 May 2025, 13:31
OCCiera wrote: ↑16 May 2025, 20:47 I keep all my music in WMA lossless
Why WMAL ? You'd better use FLAC for best compatibilty between various devices

I only use Windows devices, as I said previously, my media player is a Windows Mobile device, so it's the best option for that (I don't think it supports anything other than MP3 WAV and WMA regardless). Most of my files are stored in Windows Media Foundation formats for a few reasons.

.WDP
-Is lossless like PNG
-Supports tagging like JPEG
-Supports HDR
-Extremely well integrated into Windows' software stack

.WMA (lossless)
-dang'ol Windows CE devices mahn
-Better file sizes than FLAC (marginally)
-Also extremely well integrated into Windows' software stack

I haven't really looked into .WMV, because .MP4 is Gud Euff,^TM and I haven't really had a need to do any mass conversion like I have the other two.

Unread post by **Duke** » 19 May 2025, 02:27

K4sum1 wrote: ↑18 May 2025, 20:24 I more so meant for you to do it yourself, with music you're more familiar with.

I did it years ago, the reason why I've chosen to use q5 which is, IMHO, the best balance between file size and quality.

YouTube · Unread post by **K4sum1** » 24 May 2025, 06:27

I found this, and found it interesting.

https://www.rarewares.org/rrw/eaqual.php

Sample size of 1, but I compressed a song with both aoTuV and libvorbis, and it preferred the aoTuV file for q4, q4.5, q5, q5.5, q6, q6.5, q7, q7.5, and it was equal for q8.

128 Opus looks to score on par with a q5 aoTuV or q5.5 libvorbis conversion. Maybe Opus score could be better if had a native 44.1 mode idk. Opus also leads for all tests as expected.

Raw data if you're curious:

► Show Spoiler

The size used by the files:

► Show Spoiler

Eclipse Community