Skip to main content
Topic: Native ffmpeg AAC encoder is stable (Read 8236 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Native ffmpeg AAC encoder is stable

Reply #1
so basically

ffmpeg -i input.wav -b:a 128k output.m4a ?

Kinda shame there is no good vbr mode, since I have bunch of voice only stuff > that doesn't need 128kbps at all.
PANIC: CPU 1: Cache Error (unrecoverable - dcache data) Eframe = 0x90000000208cf3b8
NOTICE - cpu 0 didn't dump TLB, may be hung

Native ffmpeg AAC encoder is stable

Reply #2
so basically

ffmpeg -i input.wav -b:a 128k output.m4a ?

Kinda shame there is no good vbr mode, since I have bunch of voice only stuff > that doesn't need 128kbps at all.


Sounds like just the sort of thing HE-AAC and HE-AACv2 are designed for. Give them a test and see where they become "good enough" for you. Not sure if it's implemented in FFMPEG but if not, QAAC is a very good AAC implementation.

EDIT: Also try Opus, of course - but AAC wins out as far as compatibility with many devices.

Native ffmpeg AAC encoder is stable

Reply #3
@Moni, I will just keep
Code: [Select]
# encode audio with apple AAC encoder via afconvert
    afconvert -v -f "m4af" -s 3 input.aif

in my scripts (due to huge amount of laziness).
PANIC: CPU 1: Cache Error (unrecoverable - dcache data) Eframe = 0x90000000208cf3b8
NOTICE - cpu 0 didn't dump TLB, may be hung

Native ffmpeg AAC encoder is stable

Reply #4
Great news! Hopefully the FFMPEG AAC encoder reaching a decent state means it will attract more developers. I get the sense that the multiple Open Source AAC encoder implementations scattered across multiple projects and different licenses is not the most efficient use of limited talent. At least this version is under LGPL and GPL.

The only thing that is not clear to me is whether this already implements Temporal Noise Shaping or whether it's still disabled by default. It looks as if Rostislav Pehlivanov, who worked on it as part of GSoC, wasn't able to finish that in time for GSoC's end. It's unclear to me whether he managed to finish it between then and now.
Every night with my star friends / We eat caviar and drink champagne
Sniffing in the VIP area / We talk about Frank Sinatra
Do you know Frank Sinatra? / He's dead

Native ffmpeg AAC encoder is stable

Reply #5
The only thing that is not clear to me is whether this already implements Temporal Noise Shaping or whether it's still disabled by default. It looks as if Rostislav Pehlivanov, who worked on it as part of GSoC, wasn't able to finish that in time for GSoC's end. It's unclear to me whether he managed to finish it between then and now.


It's the "-aac_tns" option and it seems enabled by default, so I'd assume it's ready.

Native ffmpeg AAC encoder is stable

Reply #6
The only thing that is not clear to me is whether this already implements Temporal Noise Shaping or whether it's still disabled by default. It looks as if Rostislav Pehlivanov, who worked on it as part of GSoC, wasn't able to finish that in time for GSoC's end. It's unclear to me whether he managed to finish it between then and now.


It's the "-aac_tns" option and it seems enabled by default, so I'd assume it's ready.

That's good to hear! In that case I'd be keen to see how it performs in the next listening test.
Every night with my star friends / We eat caviar and drink champagne
Sniffing in the VIP area / We talk about Frank Sinatra
Do you know Frank Sinatra? / He's dead

Native ffmpeg AAC encoder is stable

Reply #7
Wow, thats one slow audio encoder

Using:
- Latest win64 static build from here: http://ffmpeg.zeranoe.com/builds/
- Spec wise: SSD, i7 4th-gen processor...
- Source: 16bit/44.1khz FLAC.

Anything wrong with my parameters (foobar)?
Code: [Select]
-i - -vn -sn -c:a aac -q:a 1 %d

Or is it just that slow?
Edit: Okay, so q must be between 0.1-10. 1 gives about x22 in encoding speed, ouput at ~190kbps

Native ffmpeg AAC encoder is stable

Reply #8
[quote author=o-l-a-v link=msg=912652 date=1449673421]Or is it just that slow?[/quote]
Yes, it is slow. -ab 320, core i3 - ~9x

Native ffmpeg AAC encoder is stable

Reply #9
The quality at VBR ~165kbps (-i - -vn -sn -c:a aac -q:a 0.9 %d) is awful. Not even worth ABX'ing.
CBR on the other hand (-b:a 160k) seems legit.

Native ffmpeg AAC encoder is stable

Reply #10
I just checked out the encoders.texi file on ffmpeg git: https://github.com/FFmpeg/FFmpeg/blob/maste...c/encoders.texi

Quote
@anchor{aacenc}
@section aac

Advanced Audio Coding (AAC) encoder.

This encoder is the default AAC encoder, natively implemented into FFmpeg. Its
quality is on par or better than libfdk_aac at the default bitrate of 128kbps.
This encoder also implements more options, profiles and samplerates than
other encoders (with only the AAC-HE profile pending to be implemented) so this
encoder has become the default and is the recommended choice.


Is it true that it performs better than libfdk_aac @ 128Kbps? Last time I listened to it, it was worse than FAAC.

Native ffmpeg AAC encoder is stable

Reply #11
Is it true that it performs better than libfdk_aac @ 128Kbps? Last time I listened to it, it was worse than FAAC.

The native AAC encoder recently had a lot of improvements applied, through a lot of listening tests and back and forth in https://trac.ffmpeg.org/ticket/2686 between the devs and Kamedo2.
Since FDK 128kbps was more or less where the bar was set (from what I understand/remember), they can say it probably sounds better in a lot of samples and the encoder is "optimized" for that. Further tests and feedback will allow it to be better and faster. Since the encoder is considered stable enough, they removed the experimental flag, so the users can start to more easily give feedback.

 
SimplePortal 1.0.0 RC1 © 2008-2018