HE-AAC would probably beat RealMedia acelp sipro3 @ 16 Kbit/s for music encodings as it is more geared for music encodings and said to be highly efficient. Plain AAC-LC QT encoded (16 kbit/s) sounds worse then RealMedia spiro3 for speech encodings.

Now I wonder if HE-AAC would be better then RealMedia for both speech and music at 16 kbit/s and maybe also at a lower bitrate.
Could someone with access to a HE-AAC encoder encode both a speech and music wav sample and decode it to wav (mono same sample rate) again and post it somewhere for me to judge the quality?

Just FYI: HE-AAC isn't targeted at voice coding. The MPEG4 codec for voice is CELP.

Yes, AAC is not specifically targeted at voice encoding.
My question is if an ACELP speech coder at 16 Kbit/s like rm sipro3 could be "overtaken" in quality by a coder not specifically targeted at voice. From what I've read about AAC+ and hearing AAC and OGG encodings at 16 Kbit/s it might well be possible that the "quality distance" between the acelp speech codec rm sipro3@16Kbps and AAC (I think similar quality would be achieved at AAC 20Kbps-32Kbps, not really checked)
The good thing about it is that the more "musical" aspects of voice would be preserved - resulting in better subjective quality, a reason to choose AAC+ over realmedia spiro3 16 KHz for speech encoding (and subsequently 16 KHz/16Kbit/s WMA speech encoding) and maybe also Ogg Speex and other speech ocders at 16Kbps. A related theoretical question would be where exactly the "border bitrate" and the average subjective quality of speech and musical codecs would be, is 16 Kbps or 12 Kbps? What is the lowest possible bitrate of AAC-HE? Will there be an open source encoder eventually?

About MPEG4 celp I have posted a question in the speech forum.