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: Streaming AAC (Read 7222 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Streaming AAC

Hi!

We are readying a service called Radical.FM to compete with Pandora and the like. We did one of the first interactive radio stations in the world ten years ago in Sweden. Tomsradio.com predated AAC+ and we chose WMA for various reasons. Audiograbber worked great back then, ripping CDs directly to WMA files and trimming dead air and normalizing volume all in one step.

For our relaunch we have chosen AAC+ (HE-AAC v1) VBR at 64k. It is my understanding that version 1 gives the best result between 48k and 96k (v2 below 48 and straight AAC above 96).

We want to create a lossless master library from CDs, probably in WAV or FLAC, with maximum song info attached to each file. At a legal minimum we are required to report Artist, Title, and ISRC (International Standard Recording Code) for every track we stream. More info like cover art and lyrics would be nice.

Then we’ll need to batch convert the entire library to AAC+ with the best system that will retain the attached data.

Somewhere in there we have to normalize volume across the entire library, and cut any dead air at the head or tail of tracks.

It would appear that dBpoweramp is a popular converter CD/lossless? And consensus seems to have Nero 1.5.1 leading the pack today for compression to AAC+?

A few questions;
1)   Is dBp better than Nero for converting, if so, why?
2)   Can Nero batch convert an entire lossless library to AAC+ and retain attached data, if not what can?
3)   Which can/should do the normalization and trimming?
4)   Is there a superior package or component or school of thought?

Free is nice, but nicer is being sure we have the program that does the best job. Our concern is to get our CD library to your ears in the best possible way a 64k stream will allow.

Thanks in advance for your time!
Best Regards/Tom

Streaming AAC

Reply #1 –
Foobar2000 can certainly manage the first two questions. Not entirely sure what you mean by 'trimming', but foobar can replaygain your collection then write the results directly to the files if required (I believe). I know it can with MP3 as I've done this myself.

Streaming AAC

Reply #2 –
1) Compare their output with your own ears
2) You can script everything - a skillful Linux system administrator can script that in an hour, you just need a tool to extract data from source and tag the resulting files (the former can be easily accomplished if you use FLAC, for the latter you can use faac)
3) mencoder can normalize audio, sox can trim them
4) I've no idea, I'm a UNIX guy.

Streaming AAC

Reply #3 –
The manpage for the SoX silence function says it's not reliable for trimming silence from the end of a file (it recommends reversing the file first). I use shntool for trimming digital silence from both ends of lossless files. It's not as sophisticated as SoX (it trims only zero-value samples and doesn't do anything in the middle of files) but it meets my needs. However there's no way to really automate it. I tried to automate it via foobar2000's converter interface but it's not compatible.

Quote
we have chosen AAC+ (HE-AAC v1) VBR at 64k. It is my understanding that version 1 gives the best result between 48k and 96k (v2 below 48 and straight AAC above 96).


You should test and try different settings for yourself. I have a private v1 stream @ 128 kbps and it sounds just dandy.

I was wondering why you were pre-encoding instead of doing it on the fly, but I guess if you're doing a personalized stream for each listener, it's less hassle.

Streaming AAC

Reply #4 –
Quote from: mjb2006 on
You should test and try different settings for yourself. I have a private v1 stream @ 128 kbps and it sounds just dandy.


If 128k HE sounds fine, you'd probably be happy with much lower bitrates.  128k LC is usually considered much better quality at that bitrate since it includes treble frequencies as well, and not just harmonics of the bass and mids. 

Streaming AAC

Reply #5 –
Quote from: saratoga on
If 128k HE sounds fine, you'd probably be happy with much lower bitrates.


I am happy with lower bitrates, but my streaming AAC encoder supports HE only. I can choose up to 128 kbps. I have no bandwidth shortage, so I use 128 on my LAN and 80 over the Internet, and both are quite good.

Quote from: saratoga on
128k LC is usually considered much better quality at that bitrate since it includes treble frequencies as well, and not just harmonics of the bass and mids.


I understand the technical reasons why LC is more accurate, but SBR works because most people can't perceive the kind of treble inaccuracies it produces. I'm in that category; for the audio I usually listen to, the 128k HE stream sounds so much like the 128k LC files to my ears that I'm hesitant to just accept what's "usually considered" or the OP's understanding that LC is better than HE at that bitrate and that HE is only good for lower rates. That's why I wanted to reiterate the advice to the OP to assess quality and optimal settings based on his own testing. He may find the quality tradeoffs with different combinations of bitrate & AAC profile aren't what he expects. He may also want to test different encoders, as well as what happens when an HE stream is played on a device that only supports the LC portion of the stream and how that compares to LC at the same bitrate.

Streaming AAC

Reply #6 –
Thanks for all the tips!

Obviously one can spend much time testing; but it’s still subjective. Serious professionals have done so before, and this is one of a million jobs we have, so I’d rather not reinvent the wheel. Our bandwidth costs will be ENORMOUS if we are successful; if unsuccessful then it is all academic. We will use 64k v1 for free streams and if we offer a premium service we will likely use straight 128k AAC for that. Even storage on dedicated servers is expensive so only the compressed libraries can be stored online. And yes, we are supplying individual streams to every user, so nothing but playlist generation can be done on the fly.

To keep focused, our IMMEDIATE need is simply to choose the most suitable converter from CD to WAV to get started on our master library. We have a couple weeks to pick or write the best batch encoder.

So a reiteration of the initial question; is dBp better than Nero for our conversion purposes? Is there another contender worth considering? EAC was what I expected might be suggested, and we are still considering it. I cannot yet tell if it can trim dead air at heads and tails. And I’m beginning to feel that decade old Audiograbber may STILL be our best bet with its normalization and auto-trimming capabilities. But I think it only accesses freeDB track/artist info – not sure yet how limiting that is.

Well, thanks again to everyone that bothers to read/post. Much appreciated!

Streaming AAC

Reply #7 –
Quote from: tom@redbuggy.com on
We want to create a lossless master library from CDs, probably in WAV or FLAC, with maximum song info attached to each file. At a legal minimum we are required to report Artist, Title, and ISRC (International Standard Recording Code) for every track we stream. More info like cover art and lyrics would be nice.

Then we’ll need to batch convert the entire library to AAC+ with the best system that will retain the attached data.

Somewhere in there we have to normalize volume across the entire library, and cut any dead air at the head or tail of tracks.

It would appear that dBpoweramp is a popular converter CD/lossless? And consensus seems to have Nero 1.5.1 leading the pack today for compression to AAC+?

A few questions;
1)   Is dBp better than Nero for converting, if so, why?
2)   Can Nero batch convert an entire lossless library to AAC+ and retain attached data, if not what can?
3)   Which can/should do the normalization and trimming?
4)   Is there a superior package or component or school of thought?

Free is nice, but nicer is being sure we have the program that does the best job. Our concern is to get our CD library to your ears in the best possible way a 64k stream will allow.

Thanks in advance for your time!
Best Regards/Tom


To quickly answer some of your questions with regards to dBpoweramp (I will not offer comparisons with other programs).

FLAC would be better to create your backend master library from, you get 2:1 compression and other programs can read the ID tags / Artwork, dBpoweramp can create Wave files with full ID tags, but once you step out of using dBpoweramp there is not much that would read the ID tags (on the wave files, not AAC).

ISRC is suppiled with AMGs metadata service for many discs (one of the 5 databases dBpoweramp uses for Ripping), Album art is also pulled from the internet. How many discs do you have to rip? if over 300 I would recommend a robotic loader and the Batch Ripper program (it is used by most of the top commercial ripping houses out there).

dBpoweramp has a Batch Converter, you can batch convert FLAC to AAC, normalize, trim silence all in one step with the included DSPs.
 

Streaming AAC

Reply #8 –
Quote from: tom@redbuggy.com on
Audiograbber may STILL be our best bet with its normalization and auto-trimming capabilities.


Its silence trimming is threshold based, but you can't control the threshold. Its normalization is nothing special, just the usual post-processing of the WAV to scale it up to 0 dbFS peaks, which usually is a very tiny adjustment these days. Have you considered using Replay Gain? You could tag the lossless files and leave the audio data untouched, and use AACGain to scale your untagged (as far as RG goes) AAC files for broadcast.