[FFmpeg-user] LUFS measurment for short length audio

Bouke bouke at editb.nl
Mon Sep 14 22:25:51 EEST 2020



> On 14 Sep 2020, at 15:56, christian.will at 4-real.com wrote:
> 
> 
> 
>> On 14 Sep 2020, at 14:27, christian.will at 4-real.com <mailto:christian.will at 4-real.com> wrote:
> 
>> 
>> I tried my example with adding silence before and after and i get both times -23.3db LUFS for i & m. 
> 
>>> What happens when you use -ss and -t besides the silence (to crop the silence start / end, work here…)
> 
> How does this work? Does it crop the silence from the file or is it just for not measuring the silence parts at beginning and end??
> Do you have a command i can try?

Put in -ss <duration added silence start in secs> -t <duration added silence at end in secs> after the -i <input>

Bouke

> 
> 
>> Which is a 1db variation to the customers value. If all files are coming with a 1db difference, that would be something we could work with.
>> So we are not precisely here, but its good enough for now.
>> I wonder how the meters are measuring it. As they dont have problmes with durations.
>> We will try ... 
> 
>>> 1 dB is not really a big difference I would think, and since this is not broadcast I take it, it ‘should’ not matter, as long as all input gets to the same percepted value.
> 
> Yes its not much, but its mostly due to controlling issues, not to hit the same value. So the measurement must be somehow exact, up to +-2db variation to make sure the controlling is useful. When it comes to close-up diaologs and heavy compression, 2 - 4 db can be a lot. But this is the minor case mostly 😊
> 
> 
> What are you doing, game assets?
> 😊 yep
> 
> 
> Bouke
> 
>> 
>> 
>> -----Ursprüngliche Nachricht-----
>> Von: ffmpeg-user <ffmpeg-user-bounces at ffmpeg.org> Im Auftrag von Bouke
>> Gesendet: Montag, 14. September 2020 14:08
>> An: FFmpeg user questions <ffmpeg-user at ffmpeg.org>
>> Betreff: Re: [FFmpeg-user] LUFS measurment for short length audio
>> 
>> 
>>> On 14 Sep 2020, at 13:54, <christian.will at 4-real.com> <christian.will at 4-real.com> wrote:
>>> 
>>> Its all mono we are talking about here
>>> 
>>> Full command line would be
>>> ffmpeg -nostats -i '#{filename}' -filter_complex ebur128 -f null -
>>> 2>&1 (i only forward this from the IT section, but its the standard
>>> ffmpeg lufs measurment as far as i know)
>>> 
>>> yes, we dont know if the customer values are correct. The problem is, there is no correct if you dont use the same method/algo.
>>> There are different meters coming up with different values on LUFSs, specially if you are under 0,4s.
>>> You are right, the best would be to use the same method, but our customer wont share their method and we cant reverse engineer something we dont own.
>>> 
>>> Yes, there ist the (m) value in the output, but i think its not very 
>>> reliable ... as the file measured isnt silent 😊
>> 
>> I would think M is not relevant, as it should equal I if the file is 400 Msecs… And I can reproduce the issue on a short file. But, if I cat the file with silence before / after, then measure with adding -ss yadda -t sourcedur it seems to work.
>> (But a looped input results the same values when scanned fully.)
>> 
>> 
>> 
>> Bouke.
>> 
>> 
>>> 
>>> 
>>> 
>>> -----Ursprüngliche Nachricht-----
>>> Von: ffmpeg-user <ffmpeg-user-bounces at ffmpeg.org> Im Auftrag von 
>>> Bouke
>>> Gesendet: Montag, 14. September 2020 13:41
>>> An: FFmpeg user questions <ffmpeg-user at ffmpeg.org>
>>> Betreff: Re: [FFmpeg-user] LUFS measurment for short length audio
>>> 
>>> 
>>>> On 14 Sep 2020, at 13:23, christian.will at 4-real.com wrote:
>>>> 
>>>> 
>>>> To cat means to loop the file until it is longer then 0,4s?
>>> 
>>> yes
>>> 
>>>> 
>>>> Yes we tried, there is a variation of 3db or more, so it is kind of suboptimal.
>>> 
>>> If it’s exact 3db, it smells like you might render the file to mono on catting, are you sure that’s not the case?
>>> 
>>>> We also know the loudnorm command.
>>> 
>>> Why, if you want to conform to R128 specs, you analyse, get the I 
>>> value, and lower / up the volume by the measured Integrated minus 
>>> target Integrated (in Db, one LU is equal to one dB)
>>>> 
>>>> We use this command atm to read out values:
>>>> 
>>>> ffmpeg -nostats -i 'filename' -filter_complex ebur128 -f null
>>> 
>>> This might or might not work, if it’s 5.1, you would need to omit the LFE channel, and if there are multiple tracks you would need to patch the correct ones first.
>>> So, full command line / output missing :-) (I would never think I 
>>> would write this…)
>>> 
>>>> and it is fine with all files >0,4s. As it should be relating to the ebu128 definition.
>>>> 
>>>> Anyway, our customer provides us with lufs values for shorter lenghts and he wont tell us how he analyses it.
>>> 
>>> So you have no clue if your client measurements are correct? I would not accept this, you would need to reverse engineer your clients steps...
>>> 
>>>> 
>>>> Now we tried to read out momentary lufs, as it works with 400ms windows, but im not sure if i get correct values here.
>>>> As it shows -120db mLUFS for an example and on proTools it says 
>>>> -23--24db (with DPMeterXP2)
>>> 
>>> -120 is silence…
>>> 
>>>> 
>>>> Anybody an idea how to read out reliable momentary LUFS values with ffmpeg?
>>> 
>>> Not sure how reliable it is, but it’s just in the output I would 
>>> think…
>>> 
>>> Bouke
>>> 
>>> 
>>>> 
>>>> -----Ursprüngliche Nachricht-----
>>>> Von: ffmpeg-user <ffmpeg-user-bounces at ffmpeg.org> Im Auftrag von 
>>>> Gyan Doshi
>>>> Gesendet: Montag, 14. September 2020 09:50
>>>> An: ffmpeg-user at ffmpeg.org
>>>> Betreff: Re: [FFmpeg-user] LUFS measurment for short length audio
>>>> 
>>>> 
>>>> 
>>>> On 14-09-2020 01:15 pm, Bouke wrote:
>>>>>> On 09 Sep 2020, at 10:33, christian.will at 4-real.com wrote:
>>>>>> 
>>>>>> 
>>>>>> 
>>>>>> Hi list!
>>>>>> 
>>>>>> 
>>>>>> 
>>>>>> Can somebody help to measure LUFS for audio files under 0,4 seconds???
>>>>>> 
>>>>> Cat it a couple of times first?
>>>> 
>>>> A few more times is required. Total duration should be  >3 seconds.
>>>> 
>>>> See https://superuser.com/q/1281327/
>>>> 
>>>> Gyan
>>>> _______________________________________________
>>>> 
>>> 
>>> _______________________________________________
>>> ffmpeg-user mailing list
>>> ffmpeg-user at ffmpeg.org
>>> https://ffmpeg.org/mailman/listinfo/ffmpeg-user
>>> 
>>> To unsubscribe, visit link above, or email ffmpeg-user-request at ffmpeg.org with subject "unsubscribe".
>>> 
>>> _______________________________________________
>>> ffmpeg-user mailing list
>>> ffmpeg-user at ffmpeg.org
>>> https://ffmpeg.org/mailman/listinfo/ffmpeg-user
>>> 
>>> To unsubscribe, visit link above, or email 
>>> ffmpeg-user-request at ffmpeg.org with subject "unsubscribe".
>> 
>> _______________________________________________
>> ffmpeg-user mailing list
>> ffmpeg-user at ffmpeg.org
>> https://ffmpeg.org/mailman/listinfo/ffmpeg-user
>> 
>> To unsubscribe, visit link above, or email ffmpeg-user-request at ffmpeg.org with subject "unsubscribe".
>> 
>> _______________________________________________
>> ffmpeg-user mailing list
>> ffmpeg-user at ffmpeg.org
>> https://ffmpeg.org/mailman/listinfo/ffmpeg-user
>> 
>> To unsubscribe, visit link above, or email 
>> ffmpeg-user-request at ffmpeg.org with subject "unsubscribe".
> 
> _______________________________________________
> ffmpeg-user mailing list
> ffmpeg-user at ffmpeg.org <mailto:ffmpeg-user at ffmpeg.org>
> https://ffmpeg.org/mailman/listinfo/ffmpeg-user <https://ffmpeg.org/mailman/listinfo/ffmpeg-user>
> 
> To unsubscribe, visit link above, or email ffmpeg-user-request at ffmpeg.org <mailto:ffmpeg-user-request at ffmpeg.org> with subject "unsubscribe".
> 
> _______________________________________________
> ffmpeg-user mailing list
> ffmpeg-user at ffmpeg.org <mailto:ffmpeg-user at ffmpeg.org>
> https://ffmpeg.org/mailman/listinfo/ffmpeg-user <https://ffmpeg.org/mailman/listinfo/ffmpeg-user>
> 
> To unsubscribe, visit link above, or email
> ffmpeg-user-request at ffmpeg.org <mailto:ffmpeg-user-request at ffmpeg.org> with subject "unsubscribe".



More information about the ffmpeg-user mailing list