Re: Fast(er) transcoding from aac to mp3.
On Sat, 19 Mar 2011 00:06:23 +, you wrote: Looking back at my script I wrote back at Christmas when the iphone option stopped working and my programs started coming down in aac I seem to have a faster solution which works for me; time( ffmpeg -i You_and_Yours_-_16_03_2011_b00zf33w_default.aac -map_meta_data You_and_Yours_-_16_03_2011_b00zf33w_default.mp3:You_and_Yours_-_16_03_2011_b00zf33w_default.aac You_and_Yours_-_16_03_2011_b00zf33w_default.mp3) real2m4.772s user2m2.440s sys 0m1.580s it is faster than std ffmpeg + lame b 128 but the output file is half the size (bitrate is 64 instead of 128 kb/s) using lame -b 64 is 6 secs slower than your method - 8%? (encoding a 21 minute aac) JC Std: -- FFmpeg version git-c9e16a9-Sherpya, Copyright (c) 2000-2011 the FFmpeg developer LAME 3.98.2 32bits (http://www.mp3dev.org/) Using polyphase lowpass filter, transition band: 16538 Hz - 17071 Hz Encoding stdin to Adam_Dalgliesh_-_A_Taste_for_Death_-_Episode_5_b00tb9n1_default.partial .aac.ffmpeg.new.mp3 Encoding as 44.1 kHz j-stereo MPEG-1 Layer III (11x) 128 kbps qval=3 real1m49.862s user1m47.266s sys 0m0.497s -- Without lame pipe: -- FFmpeg version git-c9e16a9-Sherpya, Copyright (c) 2000-2011 the FFmpeg developer : Output #0, mp3, to 'Adam_Dalgliesh_-_A_Taste_for_Death_-_Episode_5_b00tb9n1_defa ult.partial.ffmpeg.new.wolame.mp3': Metadata: TSSE: Lavf52.95.0 Stream #0.0: Audio: libmp3lame, 44100 Hz, stereo, s16, 64 kb/s Stream mapping: Stream #0.0 - #0.0 Press [q] to stop encoding size=9845kB time=1260.09 bitrate= 64.0kbits/s video:0kB audio:9844kB global headers:0kB muxing overhead 0.000327% real1m19.481s user0m0.000s sys 0m0.062s -- With lame -b 64 i get : LAME 3.98.2 32bits (http://www.mp3dev.org/) Resampling: input 44.1 kHz output 24 kHz Using polyphase lowpass filter, transition band: 10935 Hz - 11226 Hz Encoding stdin to Adam_Dalgliesh_-_A_Taste_for_Death_-_Episode_5_b00tb9n1_default.partial .aac.ffmpeg.new.mp3 Encoding as 24 kHz j-stereo MPEG-2 Layer III (12x) 64 kbps qval=3 real1m25.357s user1m19.810s sys 0m0.622s ___ get_iplayer mailing list get_iplayer@lists.infradead.org http://lists.infradead.org/mailman/listinfo/get_iplayer
Re: Fast(er) transcoding from aac to mp3.
dinkypumpkin wrote: I don't remember if it's the same in Lucid, but in Ubuntu Maverick I just needed to install libavcodec-extra-52 (which replaces libavcodec-52 installed by default with ffmpeg) for MP3 support. Hmm, i'll look into that. Maybe i can shoehorn it into Lenny. Thanks CJ ___ get_iplayer mailing list get_iplayer@lists.infradead.org http://lists.infradead.org/mailman/listinfo/get_iplayer
Re: Fast(er) transcoding from aac to mp3.
dinkypumpkin wrote:
Oops, forget that. I see that you're looking for a statically-linked ffmpeg.
Well that's only due to Debian problems ('oldness' and licensing issues
with mp3).
I notice that i've got libavcodec51 installed, but it's done me no good
at all.
CJ
___
get_iplayer mailing list
get_iplayer@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/get_iplayer
Fast(er) transcoding from aac to mp3.
Hi I sometimes download BBC radio shows to listen on my mp3 player. It looks like I'm going to have to start transcoding them from aac to mp3. Using Ubuntu I did some tests to see just how long it takes to transcode these things. I downloaded a 3-hour show (Steve Wright in the Afternoon*) and calculated the time taken to convert it using ffmpeg. This is the command:- time ffmpeg -i filename.aac -acodec libmp3lame -ab 128k test1.mp3 The result:- real 20m42.369s user 17m40.318s sys 0m56.812s That's over 20 minutes! During the test I wasn't running any other heavy programs. I kept checking the System Monitor, it was always up there at 98%, 99%, 100%. There's nothing wrong with my ffmpeg, it's a new one from git. Doing the best it can on my Celeron D 2.26GHz computer. Then I repeated the test using GOGO-no-coda encoder, a turbocharged version of LAME. It needs wav as input so I piped it through ffmpeg. I converted the same show using this command:- time ffmpeg -i filename.aac -f wav - | gogo -b 128 -q 0 stdin test2.mp3 The result:- real 10m30.195s user 8m12.411s sys 0m31.382s That's about half the time. If any of you feel like trying GOGO-no-coda it's available in Ubuntu's repo. Install it like this:- sudo apt-get install gogo Or download a deb from somewhere. For Windows users it's available from rarewares.org and other places. I suppose, if the programmes are talk-shows then it will be OK to reduce the q setting for gogo and also maybe reduce the bitrate to 96Kbps or 64Kbps. That would speed things up even more. If any of you have experience using different mp3 encoders it would be good to know how they compare with gogo. * I didn't listen to the show. For test purposes only. :-) ___ get_iplayer mailing list get_iplayer@lists.infradead.org http://lists.infradead.org/mailman/listinfo/get_iplayer
Re: Fast(er) transcoding from aac to mp3.
On 03/18/2011 12:52 PM, bat guano wrote: -acodec libmp3lame -ab 128k test1.mp3 Hmm. For some reason ffmpeg is not pulling in mp3lame codec when I compile. A test I did of time (ffmpeg -i You_and_Yours_-_16_03_2011_b00zf33w_default.aac -ab 128k test.wav; lame test.wav test.mp3) finished in 4:35 - this is on a core duo at 1.8GHz. So, 12* or so real-time. If doing 2 encodings at once, 24*. With vbr (lame -v) - then it goes faster, 3:24 - about 18*. Seems a fair bit faster. Recent (this week) versions of lame and ffmpeg, gcc 4.4.4. ___ get_iplayer mailing list get_iplayer@lists.infradead.org http://lists.infradead.org/mailman/listinfo/get_iplayer
Re: Fast(er) transcoding from aac to mp3.
On Fri, 18 Mar 2011 12:52:02 +, you wrote: Hi I sometimes download BBC radio shows to listen on my mp3 player. It looks like I'm going to have to start transcoding them from aac to mp3. I just joined this list yesterday - so I don't know the context of this message - Has the BBC dropped MP3 from radio broadcasts? I noticed this in the last few days. Is it a permanent switch? time ffmpeg -i filename.aac -acodec libmp3lame -ab 128kinfu test1.mp3 Does anyone have this working on Windows? I talked to Phil about 18 months ago about this. Only worked on *nix. So I use faad instead - but it is very cpu intensive. Does anyone know of a free AAC(ADTS) to AAC converter - to play on an ipod (yes I know). Better quality than AAC to MP3 conversion? JC ___ get_iplayer mailing list get_iplayer@lists.infradead.org http://lists.infradead.org/mailman/listinfo/get_iplayer
Re: Fast(er) transcoding from aac to mp3.
Not really comparing like-for-like, but with 2010 Macbook Pro (2.66 Ghz Core i7, 8Gb) and ffmpeg HEAD, transcoding a 3-hour program gives these numbers: real5m7.197s user4m56.078s sys 0m6.127s Just for kicks, in a 32-bit 1-CPU Ubuntu 10.10 VM (via VMWare Fusion) on the same machine I get this for the same file: real5m48.907s user5m30.741s sys 0m11.337s ffmpeg saturated one of the CPU cores in both cases, so the difference is more-or-less from the VM overhead. That's OK for my purposes, but it's not hard to see how it could be pokey on older hardware. On 18 Mar 2011, at 12:52, bat guano wrote: I downloaded a 3-hour show (Steve Wright in the Afternoon*) and calculated the time taken to convert it using ffmpeg. This is the command:- time ffmpeg -i filename.aac -acodec libmp3lame -ab 128k test1.mp3 The result:- real20m42.369s user17m40.318s sys0m56.812s ___ get_iplayer mailing list get_iplayer@lists.infradead.org http://lists.infradead.org/mailman/listinfo/get_iplayer
Re: Fast(er) transcoding from aac to mp3.
On 03/18/2011 03:50 PM, bat guano wrote: A test I did of time (ffmpeg -i You_and_Yours_-_16_03_2011_b00zf33w_default.aac -ab 128k test.wav; lame test.wav test.mp3) finished in 4:35 - this is on a core duo at 1.8GHz. So, 12* or so real-time. @ Ian Surely you're performing two processes there? First converting aac to wav... then converting wav to mp3. Would it not be better to use a pipe? Indeed it would, I couldn't be bothered working out how to get ffmpeg to write, and lame to read from a pipe, as I already knew the syntax to get it to do files. The time will change slightly, as the decoding and encoding canbe done on different cores, but not much, as the decoding is so enormously faster. ___ get_iplayer mailing list get_iplayer@lists.infradead.org http://lists.infradead.org/mailman/listinfo/get_iplayer
Re: Fast(er) transcoding from aac to mp3.
On 18 March 2011 14:29, James Cook james.c...@bluewin.ch wrote: Does anyone know of a free AAC(ADTS) to AAC converter - to play on an ipod (yes I know). Better quality than AAC to MP3 conversion? No need to convert, just remux using ffmpeg and the -absf aac_adtstoasc option: ffmpeg -i input.aac -vn -acodec copy -absf aac_adtstoasc output.mp4 The latest bundled ffmpeg in mplayer-win32 handles this: http://sourceforge.net/projects/mplayer-win32/files/FFmpeg/git-c9e16a9/ HTH Shevek ___ get_iplayer mailing list get_iplayer@lists.infradead.org http://lists.infradead.org/mailman/listinfo/get_iplayer
Re: Fast(er) transcoding from aac to mp3.
On Fri, 18 Mar 2011 15:50:38 +, you wrote: Like this:- time ffmpeg -i filename.aac -f wav - | lame -b 128 - test3.mp3 To my surprise the above does work on windows/cygwin I have been using faad -w aacfile | lame ... up to now - for historical reasons - (SBR not implemented error) see http://linuxcentre.net/bbc-iplayer-aac-radio-streams-now-available-using-get_iplayer . Anyway ... I just did a rough experiment and faad seems to be about 2 times slower than ffmpeg. BTW I use lame --abr 56 for spoken word programs - the files come out half as big as with -b 128. JC ___ get_iplayer mailing list get_iplayer@lists.infradead.org http://lists.infradead.org/mailman/listinfo/get_iplayer
Re: Fast(er) transcoding from aac to mp3.
On Fri, 18 Mar 2011 18:54:03 +, you wrote: No need to convert, just remux using ffmpeg and the -absf aac_adtstoasc option: ffmpeg -i input.aac -vn -acodec copy -absf aac_adtstoasc output.mp4 The latest bundled ffmpeg in mplayer-win32 handles this: http://sourceforge.net/projects/mplayer-win32/files/FFmpeg/git-c9e16a9/ HTH Shevek Thanks for your message, I tried it and it works! Now I just have to figure how to tag aac files JC ___ get_iplayer mailing list get_iplayer@lists.infradead.org http://lists.infradead.org/mailman/listinfo/get_iplayer
Re: Fast(er) transcoding from aac to mp3.
Looking back at my script I wrote back at Christmas when the iphone option stopped working and my programs started coming down in aac I seem to have a faster solution which works for me; time( ffmpeg -i You_and_Yours_-_16_03_2011_b00zf33w_default.aac -map_meta_data You_and_Yours_-_16_03_2011_b00zf33w_default.mp3:You_and_Yours_-_16_03_2011_b00zf33w_default.aac You_and_Yours_-_16_03_2011_b00zf33w_default.mp3) real2m4.772s user2m2.440s sys 0m1.580s WRT to piping; I would guess that piping output of aac - wav to wav - mp3 is marginally faster than writing to a file and then reading from a file but surely it's faster to do everything in one go? Sy On 18/03/2011 16:22, Ian Stirling wrote: On 03/18/2011 03:50 PM, bat guano wrote: A test I did of time (ffmpeg -i You_and_Yours_-_16_03_2011_b00zf33w_default.aac -ab 128k test.wav; lame test.wav test.mp3) finished in 4:35 - this is on a core duo at 1.8GHz. So, 12* or so real-time. @ Ian Surely you're performing two processes there? First converting aac to wav... then converting wav to mp3. Would it not be better to use a pipe? Indeed it would, I couldn't be bothered working out how to get ffmpeg to write, and lame to read from a pipe, as I already knew the syntax to get it to do files. The time will change slightly, as the decoding and encoding canbe done on different cores, but not much, as the decoding is so enormously faster. ___ get_iplayer mailing list get_iplayer@lists.infradead.org http://lists.infradead.org/mailman/listinfo/get_iplayer ___ get_iplayer mailing list get_iplayer@lists.infradead.org http://lists.infradead.org/mailman/listinfo/get_iplayer
