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: exhale - Open Source USAC encoder (Read 183010 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Re: exhale - Open Source USAC encoder

Reply #1075
Hello,
There's a song that has instruments at 11.25khz and exhale seems to only fill individual frequencies up to 11khz so it is removing the instrument.. is there a way to set the cutoff for filling in individual frequencies higher?
Thanks

Re: exhale - Open Source USAC encoder

Reply #1076
Hi,

which of exhale's bit-rate presets are you using? And which sampling rate is your input audio at? If you could post a few seconds of your audio file, that may also help.

Chris
If I don't reply to your reply, it means I agree with you.

Re: exhale - Open Source USAC encoder

Reply #1077
Preset B and Preset C
Both seem to have pretty similar tuning (at least, looking at a spectrogram)... relatively accurate frequencies up to 11khz and then it  goes really, really inaccurate to the point where an instrument isn't audible in the audio anymore or it just outright removes it
I would rather it go up to 11.25 or 11.5khz and spend an extra kbps or two.. if there's an instrument it is cutting off
Audio is 44.1khz

Re: exhale - Open Source USAC encoder

Reply #1078
That's a limitation of exhale's SBR encoder. Try using preset 2 or 3 instead of the letters, or resample your input to 48 kHz before encoding.

Chris
If I don't reply to your reply, it means I agree with you.

Re: exhale - Open Source USAC encoder

Reply #1079
Thanks!
Resampling to 48khz brought back some of the instrument
It's not nearly as audible as the original audio source but at least it's partially there
Looks like it raised the frequency cutoff about 500-750hz, nice
Interestingly, 48khz is 55kbps and 44.1khz is 57kbps...


Re: exhale - Open Source USAC encoder

Reply #1081
Привет!
From 12-25 seconds, cymbal suppression is noticeable.
> Download

Re: exhale - Open Source USAC encoder

Reply #1082
can someone help me im trying to play xhe audio on my galaxy s7 but it wont recognize the files, how do i make it work

Re: exhale - Open Source USAC encoder

Reply #1083
can someone help me im trying to play xhe audio on my galaxy s7
You can't. Because the lastest update for the S7 is running Android 8 and you need at least Android 9 for xHE-AAC playback.

Re: exhale - Open Source USAC encoder

Reply #1084
Exhale seems to be having an especially hard time encoding this... It generally encodes pianos pretty well but somehow not this... the bitrate is also a little high (averaged 73kbps for preset b)
preset b does sound better than preset a but it's still more artifacting than I'd expect.. I'm pretty sure even preset d
It kinda sounds wavy/ping-pong like
It seems to be the high pitched instrument causing the really high bitrate
Original is 44.1khz, I upsampled to 48khz

I guess.. the level of artifacting isn't that unusual for the presets used... I am still curious why the high pitched instrument is causing the bitrate spike
ohhhh.. I just looked at a spectrogram.. I think I see why.. there's a lot of frequencies to fill in?

Anyway, it does seem like I was being over sensitive to the artifacting on the piano... I can hear the strings broken up a bit as well.. but I don't think the level of artifacting is unusual for the presets used, I guess the piano artifacting just seemed especially annoying to me
And the bitrate spike seems to be because there's a LOT of 7khz+ frequencies to fill in.. so, mystery solved

Re: exhale - Open Source USAC encoder

Reply #1085
Indeed, that audio snippet is a tough one to encode. But I find that, compared to e.g. Winamp's HE-AAC encoder at the same bit-rate (72 kbps for comparison with preset b), exhale still sounds quite acceptable.

Thanks for sharing that sample!

Chris
If I don't reply to your reply, it means I agree with you.

Re: exhale - Open Source USAC encoder

Reply #1086
Can't tell if Fraunhofer xHE-AAC or exhale encodes it better but both are really good! I think they are tuned quite differently (Fraunhofer seems to retain the higher frequencies on this sample better to my ears)
I also find LC-AAC exhale might win against Apple in a good amount of cases?
Then again, I think Musepack at 90kbps setting might of won against 96kbps Apple AAC on this sample.. at least, with a few quick listens (Musepack averaged 130kbps, Apple AAC 135kbps)
But yeah, all the encoders averaged high.. I think Opus may have actually averaged a bit more sane (would have to check again)
I think Opus does really well when there's a lot of noisy frequencies to encode so I could see how it'd do well on this
Well, I guess the loud 7khz+ frequencies aren't tripping up Opus as much so it gains a lot of efficiency there

Re: exhale - Open Source USAC encoder

Reply #1087
Exhale seems to be adding "scratchy" sounds to this section
Even by preset c it's still quite audible...
Especially at about 0:09... for a split second it has massive artifacting
wait, I hear the same sound in the uncompressed audio... but it seems to be amplifying it?
It seems to be duplicating it... even by preset d it's quite audible, albeit a bit less
Although most of the really audible scratchy sounds disappear by preset d... it's pretty faint by that point to where you really have to listen

Re: exhale - Open Source USAC encoder

Reply #1088
Is it just my imagination or does Exhale have some slight underwater/muffled distortion and weirdness with the stereo compared to other HE-AAC encoders?  It's pretty minor but now that I notice it, it bugs me a little... I don't think it's just my imagination
It almost sounds like the stereo is slightly widened from the original audio?
I wonder if the muffled sound is because it's removing a lot of loud 2khz+ frequencies and replacing with quieter ones

Re: exhale - Open Source USAC encoder

Reply #1089
It's probably just the distortion I'm hearing and not actually something weird going on with the stereo imaging but... yeah, it just sounds "wider" to me than the uncompressed file
I'd still pick Exhale over FDK, Apple and Opus... Opus adds way too much noise and what it does to the very low frequencies is baffling (multiplies the volume), FDK has horrible ringing due to SBR kicking in really soon, Apple is really broken up.. Exhale doesn't have the issues of FDK or Apple so even with the slight distortion, I'd say it's the best

Re: exhale - Open Source USAC encoder

Reply #1090
It's probably just the distortion I'm hearing and not actually something weird going on with the stereo imaging but... yeah, it just sounds "wider" to me than the uncompressed file
Well observed. At low bit-rates and low frequencies, this may indeed happen, since there's a lot of quantization noise which may not be the same in both channels. At higher frequencies, I integrated a stereo image narrowing into exhale which should counteract this.

Chris
If I don't reply to your reply, it means I agree with you.

Re: exhale - Open Source USAC encoder

Reply #1091
Could be it, maybe
I've found c and d to be the best lower bitrate presets, b everything is a bit too broken up, d has a lot more high frequency detail and sounds slightly less "underwater" to me but c isn't bad

Re: exhale - Open Source USAC encoder

Reply #1092
I'm having some... strange behavior...
I took about 17 seconds of audio for a sample... but then I wanted a couple seconds cut off at the end so saved another sample that is about 15 seconds.. but I encoded both... the artifacting is in different areas...
on the 17 second sample, one of the loudest clicks is at about 0:02.. but I hear it in multiple areas... on the 15 second sample, the click at 0:02 is gone... but I hear some in other areas (like at 0:13 I hear multiple clicks close together on both encodes)
Also, preset a - 54kb, preset b - 62kb, preset c - 66kb

I even hear clicks by preset g... although, it's very, very faint (I of course realize you generally shouldn't use SBR this high, it's just for testing)
I think these artifacts are similar to what I was experiencing with the other sample I posted?  Maybe it's due to SBR..
okay, I posted a section in the spectrogram where a click happens... seems to be a patch of noise it adds... still pretty strange.. especially since the artifacting is different on a sample that is 2 seconds longer with nothing changed but that the sample is longer...

Nah, not exclusive to SBR... I'm hearing clicks with SBR disabled 64kbps setting as well
It's just patches of noise it is encoding...

Re: exhale - Open Source USAC encoder

Reply #1093
okay, it's possible I messed something up the first time because I just took the exact same audio sections again just to make sure and now the artifacts match up between the shorter and longer encode... so never mind, sorry

Re: exhale - Open Source USAC encoder

Reply #1094
sorry for so many posts but I figured out the difference (and it's what I suspected might be it)

One I dragged the original FLAC into Audacity and selected the time I wanted and exported... the other I selected the time, opened a new Audacity window and pasted it in... that removes the metadata... so perhaps that's it?
By pasting in a new window, it somehow removes a click or two that I hear
HOWEVER, when I do the same method with both the shorter and longer sample, there is different artifacting I'm hearing now... one, there's a click within the first second of audio... the other there isn't... I don't know if it's the metadata messing with it or if compressing a different amount of audio changes what the encoder decides to do but...

wait, now I'm very confused... no wait... both had the metadata in them... when I tested again, metadata is still there...
I'm confusing myself.. I might figure this out tomorrow but for now I just need to be done

Re: exhale - Open Source USAC encoder

Reply #1095
okay, I just did some quick testing.. I think it's Audacity
Seems to randomize something when you paste a section of audio into a new window (file size keeps changing very slightly).. I don't know if this is what is causing the differences but... yeah
Probably dithering or something
Dragging an audio file directly in, however, produces the exact same file size each time
But since the files I posted have metadata, I believe this means I did NOT copy a section and paste it into a new window.. so does seem like length of encoding DOES change what artifacts are caused where
It does seem like even the smallest changes can cause drastic artifacts in different areas.. I do wonder why these "clicks" happen... but... they do

Re: exhale - Open Source USAC encoder

Reply #1096
Eh, I guess that just happens with audio compression sometimes.. I just compressed something with 80kbps Apple LC-AAC... literally sounds like fireworks going off in some areas XD
And even the TINIEST of changes I guess can cause DRASTIC differences, it appears
Also the audio sample isn't a typical sample.. it has a few tones with a lot of blank space in-between and frequency doesn't go very high
Still a little strange but if I knew what was happening I'm sure it'd make sense
I guess it's "quantization noise" and it can just spontaneously create a massive amount of noise when bitrate is starved
Sorry for the wall of text, if I could edit previous posts I would

Re: exhale - Open Source USAC encoder

Reply #1097
Not sure I can follow everything you wrote, but thanks for clearing it up. Also note that, to avoid artifacts from, possibly, appearing at different time positions when cutting off leading samples of a waveform, you need to cut off an integer multiple of 2048 samples per channel. That's due to the (xHE-)AAC framing during encoding.

Chris
If I don't reply to your reply, it means I agree with you.

Re: exhale - Open Source USAC encoder

Reply #1098
Apple Books now accepts content encoded in xHE-AAC.

 

Re: exhale - Open Source USAC encoder

Reply #1099
I've noticed Exhale adds echo noise a lot... I hear a noise kind of like clapping
With Apple and Opus, it's more like an "air" sound.. but with Exhale it's kind of like clapping
But a lot of encoders seem to be adding some form of noise
I think it's because there's a lot of notes that don't have much sound around them... and encoders really don't like that