is it possible to encode Apple HE-AAC v2 on Windows?

[geextah_2]

Hi, I would like to know if it's possible to use Apple HE-AAC v2 on Windows? And how to achieve that?

Thanks for any help

[VideoGeek]

Sure, just download "EXHALE" app from VideoHelp---> https://www.videohelp.com/software/exhale

(also available for Linux as well)

[geextah_2]

1. exhale is not apple
2. exhale uses xHE-AAC (USAC) and Low Complexity LC-AAC, not HE v2

[davexnet]

This is the info from Duckduckgo search assist:
Using Apple's HE-AAC v2 Encoder on Windows
Installation Requirements

To use Apple's HE-AAC v2 encoder on Windows, you need to install the Apple Application Support package. This package is included with the iTunes installation but can also be extracted separately.
Steps to Install Apple Application Support

Download iTunes: Obtain the iTunes installer from the Apple website.

Extract the Installer:
Use an archive extractor (like PeaZip or 7-Zip) to open the iTunes installer without running it.

Locate and extract the AppleApplicationSupport.msi file.

Install Apple Application Support:
Run the extracted AppleApplicationSupport.msi file to install the necessary components.

Accessing the Encoder

Once the Apple Application Support is installed, applications like FFmpeg can access the CoreAudio AAC encoder, enabling the use of HE-AAC v2 for encoding audio files.

[ProWo]

This is the info from Duckduckgo search assist:
Using Apple's HE-AAC v2 Encoder on Windows
Installation Requirements
To use Apple's HE-AAC v2 encoder on Windows, you need to install the Apple Application Support package.

You can also use this way:

Open an admin terminal, write

winget install -e --id Apple.AppleApplicationSupport.x64

[geextah_2]

ok and what do I do then?

[ProWo]

ok and what do I do then?

OBS and fre:ac are supposed to recognize the Apple Core AAC encoder.
However, I haven't tested this.

[pandy]

https://github.com/tmkk/fhgaacenc

[Roph]

Why would you want to? Both Nero and FDK (with afterburner enabled) are better than Apple's encoder at HE-AACv2.

[cholla]

@ geextah_2,
I'm posting this in this thread instead of the ffmpeg 7.0.1 and Windows 7 thread.
This is from MediaInfo.Is this the conversion you are looking for?

Audio
ID                                       : 1
Format                                   : AAC LC SBR PS
Format/Info                              : Advanced Audio Codec Low Complexity with Spectral Band Replication and Parametric Stereo
Commercial name                          : HE-AACv2
Format settings                          : Explicit
Codec ID                                 : mp4a-40-2
Duration                                 : 7 min 35 s
Bit rate mode                            : Variable
Bit rate                                 : 36.7 kb/s
Maximum bit rate                         : 44.6 kb/s
Channel(s)                               : 2 channels
Channel layout                           : L R
Sampling rate                            : 44.1 kHz
Frame rate                               : 21.533 FPS (2048 SPF)
Compression mode                         : Lossy
Stream size                              : 1.99 MiB (98%)
Language                                 : English
Encoded date                             : 2026-11-07 17:41:13 UTC
Tagged date                              : 2026-11-07 17:41:21 UTC

If so the code I used was this:

ffmpeg -i input.mp4 -vn -c:a pcm_s16le -ac 2 -ar 44100 -f wav - | fhgaacenc --vbr 1 --ignorelength - output.m4a -

You need to add the path to fhgaacenc or have it in Environment Variables/Path .

[Roph]

FHG is obsolete, and is worse than both Nero and FDK (also made by fraunhofer) at HE-AACv2.

FDK support is iffy with ffmpeg due to licensing/patent concerns, but you can compile ffmpeg with FDK AAC support or just find a binary. Alternatively, Nero's encoder is fantastic at HE-AACv2.

[pandy]

If so the code I used was this:

ffmpeg -i input.mp4 -vn -c:a pcm_s16le -ac 2 -ar 44100 -f wav - | fhgaacenc --vbr 1 --ignorelength - output.m4a -

You need to add the path to fhgaacenc or have it in Environment Variables/Path .

You can remove "-" at the end - it is error - i already corrected this in my reply https://forum.videohelp.com/threads/414767-ffmpeg-7-0-1-and-Windows-7/page7#post2806246

[pandy]

FHG is obsolete, and is worse than both Nero and FDK (also made by fraunhofer) at HE-AACv2.

FDK support is iffy with ffmpeg due to licensing/patent concerns, but you can compile ffmpeg with FDK AAC support or just find a binary. Alternatively, Nero's encoder is fantastic at HE-AACv2.

Nope - hydrogen quality test rank higher FhG than FDK - in fact FDK is integer, mobile market (embedded) target - FhG is higher quality and use float instead integers.

AAC encoders ranking is pretty obvious: Apple AAC then FhG - check yourself https://wiki.hydrogenaudio.org/index.php?title=AAC_encoders and stop spreading incorrect information's.

[Roph]

There's no apostrophe for "informations".

I'm aware of that Hydrogenaudio page, yes. You may not be aware that it was last updated over 10 years ago. Check it yourself. You may have also forgotten that OP is asking about HE-AACv2.

So yes, FHG is obsolete abandonware. Winamp itself, which FHG is for, does not even exist any more. Yes FHG can indeed encode HE-AACv2, but it's a joke compared to FDK or Nero. Try it for yourself at 22, 24, 27 or 32kbit/s. It's nowhere close.

FDK *is* updated, including with a new "afterburner" option that greatly improves quality, especially for HE-AAC or HE-AACv2.

Despite being so old, Nero excels at HE-AACv2 around 20-32kbit/s. At ~32kbit/s, I'd consider Nero and FDK equal. below 28, Nero wins. Again, HE-AACv2, not LC which is what Hydrogenaudio is comparing.

It's also worth noting that this year (2026) FFMpeg's own AAC encoder got a rewrite for 9.1, though I haven't looked into it as FDK/Nero are totally satisfactory, and AAC is arguably being supplanted by Opus.

[pandy]

Once again - foundation of encoding is source - if your encoder support only 16 bit integer then in unavoidable way you reducing quality by approx 6..12cdB - if you reencode your source internally coded as constant function between 0..1 (or -0.5..0.5) then if your encoder accept only 16 bit integer you need to requantize your source from float to int - you need to perform dithering to linearize quantizer, by using TPDF dither you need at least 2 bits for dither itself i.e. you are raising noise floor by 12dB (1 bit=6.0206 dBFS), if your requantization process involve noiseshaping to increase perceived signal dynamics and match ATH then it is even worse. Noise add entropy to encoding process so reduce compression efficiency or reduce compression quality.

Hope this is more clear now - if your encoder is unable to accept at the input float or high bithdepth integers then will be unable to deliver better results - there is no magic here.

Updated ffmpeg native AAC encoder is discussed here - there is general opinion that it can't deliver quality as good as Apple AAC encoder .

FDK is tailored for mobile (embedded) devices with focus on realtime encoding - it is limited by accepting only 16 bit integer at the input.

All i wrote is provided with focus on acceptable audio quality - AAC audio with bitrate less than 64kbps in my opinion have no acceptable quality.

[Roph]

Sure it can, these files are approximately 26-27kbit/s from Nero, very impressive.

[cholla]

All of you have better hearing than I do.

As even the native AAC FFMPEG has sounds OK to my ears.
Same for if I listen to a .wav at 16 bit integer or 32 bit float.

I have some old Nero software but I have not tried it.
I don't know how to get Nero to do HE-AACv2.
I have FFMPEG set up to use Apple AAC. Only because it is supposed to be better than FFMPEG native aac.
I will need to test to see if it will do HE-AACv2
.
I worked with WinAmp's fhgaacenc because geextah posted this in the other thread.
I still use Winamp & I like it.

is it possible to use FhG AAC (Winamp) "enc_fhgaac.dll" with ffmpeg?

[pandy]

Sure it can, these files are approximately 26-27kbit/s from Nero, very impressive.

Try with more complex music - i'm old but not duff - still can hear 16kHz.

[cholla]

Here are a few things I have found trying this:

The Apple aac_at encoder will NOT encode to HE-AAC v2 on Windows.
It will encode to HE-AAC on Windows. You need the Apple Application Support package per davexnet post #4.
I did this
I also added the QT files to inside my ffmpeg/bin folder.
https://github.com/AnimMouse/QTFiles/releases

Try this:

ffmpeg -i input.wav -c:a aac_at -profile:a 4 output.m4a

To get a HE-AAC v2 You have to use libfdk_aac.

ffmpeg -i input.wav -c:a libfdk_aac -profile:a 28 output.m4a

I did a couple with Nero also.
McgeeHE.m4a & McgeeHEv2.m4a

neroAacEnc -q 0.99 -he -if input.wav -of output.m4a

neroAacEnc -q 0.99 -hev2 -if input.wav -of output.m4a

Here are the resulting files:

[cholla]

Sure it can, these files are approximately 26-27kbit/s from Nero, very impressive.

Try with more complex music - i'm old but not duff - still can hear 16kHz.

It depends on which test you use.
I tried a few & some I could not hear at 16kHz.
I could only hear to 15kHz on the site below with the individual tone test.
The multiple tone test I could hear something all the way to the end.
WARNING on the multiple test have the headphones away from your ears.
Then as the sound gets to a lower volume put them on as normal.

https://onlinesound.net/download/audio/treble-tone-freq-14khz
https://onlinesound.net/hearing-test

[pandy]

It depends on which test you use.
I tried a few & some I could not hear at 16kHz.
I could only hear to 15kHz on the site below with the individual tone test.

Well, there is many factors involved in hearing ability but there is obvious difference between half of spectrum and full spectrum - if i can compare 24 bit FLAC source and then HEAACv2 then difference is obvious, lack of space, dull, muddy sound, something nonlinear - i have nothing against audio movie track compressed to mobile use on wireless headphones where lossy compression is unavoidable - problem is music to listen stationary - this is simple no go to me.
You can't enlarge your room by placing mirrors on the wall - it may looks bigger but size is the same.

[cholla]

@ pandy ,
I was comparing oranges to oranges. Comparing the .m4as converted from a 16-bit .wav file.

I rarely have 32-bit float .wav files & maybe I could hear a difference on a good sound system.
On wireless headphones they sound close to my ears.

Even comparing the m4as to the 16-bit .wav I can definitely hear the difference but that is comparing apples to oranges.

For the hearing test site those are just to see if the sound can be heard by a person in those ranges.

[pandy]

@ pandy ,
I was comparing oranges to oranges. Comparing the .m4as converted from a 16-bit .wav file.

I rarely have 32-bit float .wav files & maybe I could hear a difference on a good sound system.
On wireless headphones they sound close to my ears.

Even comparing the m4as to the 16-bit .wav I can definitely hear the difference but that is comparing apples to oranges.

For the hearing test site those are just to see if the sound can be heard by a person in those ranges.

This is true however not in case when your source is lossy compressed (i.e. common situation for most of the sources) - if your source is lossy compressed then in compressed form is float representation (as i pointed earlier from 0..1 or -0.5..0.5) - if your lossy decoder produce at the output 16 bit quantized signal then you have or quantization errors (they sound awful and since beginning of 90's we are lucky to learn that adding dither linearize quantizer at the cost of added noise but uncorrelated with signal noise is good for overall audio quality) or signal with added noise (dither) and recommended TPDF dither has level 2LSB i.e. reduce SNR by 12dB also you may have dither with noise shaping and rarely noise shaping - noise shaping make SNR even worse but subjectively improve overall audio quality by pushing quantization errors to hearing areas where human are less sensitive (this is called psycho-acoustic) - if you take this all you see why scientifically correct is to use float representation to avoid additional requantization of signal after decoding from lossy representation. If your source is losslessly compressed or it is uncompressed then integer representation is valid however 32 bit float offer quality comparable to 23 bit integer. This shows that float is more flexible as it cover wider area.

Wireless headphones are another problem preventing to compare oranges with oranges or apples with apples as to fit audio with typical bitrate 1.5..2Mbps in BT (offering transmission speed lower than 1Mbps) you need to compress audio at least 2 but usually 4..6 times (for HQ lossy codecs) and even more so efficiently you make everything worse.
This is not a problem when you consuming music but it begin to be problem when your goal is to listen music i.e. you intentionally spent your time to listen - music is not auxiliary background.

I wish to be clear - i have nothing against HE-AAC or other comparable codec - they have own purpose and own target - in city noise with typical level 55..65dB you probably not hear too much problems with such codecs but when you listening music on good audio equipment in your home in optimal conditions then this is no go. Hope it is clear.