Recherche avancée

Sur d’autres sites (6842)

• ffmpeg : Apply a time-based expression to control lowpass, highpass, aphaser, or aecho [closed]

  17 mai 2024, par Wes Modes

  I am constructing complexFilters to procedurally generate audio with ffmpeg via fluent-ffmpeg. The interesting part of my fluent-ffmpeg complex filters look like :

  


  {
   "filter": "volume",
   "options": {
       "volume": "min(1, max(0, ((cos(PI * t * 1 / 13) * 1 + cos(PI * t * 1 / 7) * 0.5 + cos(PI * t * 1 / 3) * 0.25) +
-0.5 ) * 0.75 * -1 + 0.5))",
       "eval": "frame"
    },
    "inputs": "track_1_output",
    "outputs": "track_1_adjusted"
},


  


  Note that I am using an expression to determine the volume dynamically. To simplify my tests, I'm constructing complexFilters on the command line to understand the mysteries of some of the lightly-documented ffmpeg filters. Here's a working CLI test of using the volume filter to fade the sources in and out according to a time-based expresssion :

  


  # working fade in/out
ffmpeg -i knox.mp3 -i static.mp3 -filter_complex \
"[0:a] \
    volume = 'min(1, max(0, 0.5 + 0.5 * cos(PI * t / 5)))': eval=frame \
    [a0]; \
[1:a] \
    volume = 'min(1, max(0, 0.5 - 0.5 * cos(PI * t / 5)))': eval=frame \
    [a1]; \
[a0][a1]
    amix = inputs=2: duration=shortest" \
-c:a libmp3lame -q:a 2 output.mp3


  


  Now I want to apply lowpass, highpass, aphaser, and/or aecho to the transitions using a time-based expression. For instance, looking at lowpasss, according to ffmpeg -filters it says that it has time-based support :

  


    T.. = Timeline support
 TSC lowpass           A->A       Apply a low-pass filter with 3dB point frequency.


  


  How can I similarly apply a time-based expression to lowpass, highpass, aphaser, or aecho ? I thought the mix option might be the key, but I couldn't construct a working example and couldn't find any examples.

  


  # applying a lowpass filter
ffmpeg -i knox.mp3 -filter_complex \
"[0:a] \
    lowpass=f=3000:mix='0.5 + 0.5 * cos(PI * t / 5)';" \
-c:a libmp3lame -q:a 2 output.mp3


  


  Or if you prefer the fluent-ffmpeg filter :

  


  const ffmpeg = require('fluent-ffmpeg');

ffmpeg('knox.mp3')
    .complexFilter([
        {
            filter: 'lowpass',
            options: {
                f: 3000,
                mix: '0.5 + 0.5 * cos(PI * t / 5)'
            },
            inputs: '[0:a]',
        }
    ])
    .audioCodec('libmp3lame')
    .audioQuality(2)
    .save('output.mp3')
    .on('end', () => {
        console.log('Processing finished successfully');
    })
    .on('error', (err) => {
        console.error('Error: ' + err.message);
    });


  


  With the resultant error :

  


  [Parsed_lowpass_1 @ 0x7f9a72005c00] [Eval @ 0x7ff7bec423d8] Undefined constant or missing '(' in 't/5)'
[Parsed_lowpass_1 @ 0x7f9a72005c00] Unable to parse option value "0.5 + 0.5 * cos(PI * t / 5)"
Error applying option 'mix' to filter 'lowpass': Invalid argument


  


  What complicated fffmpeg faerie magic is needed to make some of the filters other than volume controlled by a time-based expr ?

  


• how to use ffmpeg to apply 1:1 SAR before concat on large complex filter

  24 avril 2019, par 2c2c

  I use ffmpeg to concat videos in a fashion similar to this :

  I ran into a weird error with my inputs

  [Parsed_concat_0 @ 000000002a05bb80] Input link in10:v0 parameters (size 1280x720, SAR 2049:2048) 
  
  do not match the corresponding output link in0:v0 parameters (1280x720, SAR 1:1)

  From what I’ve researched I need to use setsar to force all the videos to be 1:1 before I concat, but I’m not sure how to do that in my filter.

• avfilter/vf_convolution : add float user_rdiv[4] to allow user options to apply correctly

  24 février 2024, par Stone Chen

  avfilter/vf_convolution : add float user_rdiv[4] to allow user options to apply correctly
  Previously to support dynamic reconfigurations of the matrix string (e.g. 0m),
  
  the rdiv values would always be cleared to 0.f, causing the rdiv to be
  
  recalculated based on the new filter. This however had the side effect of
  
  always ignoring user specified rdiv values.
  Instead float user_rdiv[0] is added to ConvolutionContext which will store the
  
  user specified rdiv values. Then the original rdiv array will store either the
  
  user_rdiv or the automatically calculated 1/sum.
  This fixes trac ticket #10294, #10867.
  Signed-off-by : Stone Chen <chen.stonechen@gmail.com>
  
  Signed-off-by : Marton Balint <cus@passwd.hu>
  

  • [DH] libavfilter/convolution.h
  • [DH] libavfilter/vf_convolution.c