FFMPEG & Filter Complex: A Visual Guide to the Filtergraph Usage

[00:00] hello this is Steph and welcome to my

[00:02] channel again and today I'm going to

[00:04] show you how to use the complex filters

[00:07] uh better the filter complex using FM

[00:12] pack first of all what are the filter

[00:14] complex

[00:15] um the filter is complex are the complex

[00:18] filters filters complex is the options

[00:20] that you have in a pack is uh that

[00:23] category of filters that require uh

[00:26] multiple inputs or multiple outputs

[00:30] and we saw something in the past for

[00:32] example the concatenation for example

[00:35] with the GIF we saw the split so we have

[00:39] a rough idea about what a complex filter

[00:42] is and now we're going to see something

[00:45] a bit better in detail

[00:48] uh here we have Windows ready behind me

[00:52] at the moment it's time to get started

[01:03] okay here we are we have the terminal at

[01:06] the top uh this is the windows terminal

[01:09] so I changed from the CMD uh to this

[01:13] Windows terminal it's super cool because

[01:15] uh what you can do is also to

[01:18] select the one that you want so not only

[01:21] the command prompt that I'll be using

[01:23] most of the times but the windows

[01:25] Powershell develop compound prompt for

[01:27] visual studio

[01:30] um the

[01:32] um

[01:33] WSL windows subsystem for Linux I

[01:36] install Ubuntu I have something with a

[01:39] anaconda mini counter actually

[01:42] um like to switch from one in the other

[01:46] or having multiple tabs here it is a

[01:49] super super cool and uh

[01:53] I suggested if you would like to go on

[01:57] um coding and using the command line so

[01:59] here we have two videos

[02:02] here on the left

[02:05] where

[02:07] so videos about Singapore every people

[02:09] who follow me on Instagram uh I take

[02:12] advantage to

[02:14] give the my references about the social

[02:17] media here

[02:19] um

[02:20] so um I've been to Singapore recently

[02:23] and I would like to make a video about

[02:25] Singapore I have a couple of videos here

[02:28] play

[02:31] Singapore one and this is uh the first

[02:35] video

[02:38] Singapore Marina Bay and the marina here

[02:44] nice place to be and on the other we

[02:49] have

[02:50] Singapore

[02:55] night

[02:57] It's a Wonderful view accelerated view

[03:00] of Singapore where the

[03:06] cool so we would like to make a

[03:08] Crossfade so it's it is a transition

[03:10] from a video to another but in a linear

[03:16] way so we don't want an Abrupt change

[03:18] from one to the other it's not a

[03:19] concatenation even if we did a

[03:21] concatenation

[03:22] advantage to show you

[03:25] part about the concatenation here we use

[03:28] the a filter complex actually but what

[03:30] we would like to use is a so-called

[03:32] Crossfade so here we have a

[03:36] the

[03:37] um the Crossfade filter there is a

[03:40] called The X fade so the base code

[03:43] stayed like this

[03:45] and

[03:48] what you would like to use is a F10 pack

[03:51] minus I and the filter the freight and

[03:55] the Fellas we want this input so here if

[03:59] from Peg minus I single

[04:03] one and then minus I single port two

[04:07] these are the input files and then we

[04:09] are immediately using the filter complex

[04:12] so the filter complex is uh something

[04:15] that accepts uh

[04:17] multiple inputs multiple outputs and you

[04:20] can also apply multiple filters here so

[04:24] here we have for example the first input

[04:28] and the second input so here we can put

[04:31] zero and one like this so the input is

[04:35] uh put it this way before the name of

[04:38] the field name of the and

[04:44] after

[04:46] the equal sign we got to put the filter

[04:49] configuration so here we've got to put

[04:51] the parameter transition

[04:55] here my transition I want to be a Fades

[04:58] it's fading from a video to another

[05:01] I got to set the duration of the fade in

[05:04] this case it is one second

[05:07] and also

[05:10] um we got to put the offset what is the

[05:13] offset the offset is the

[05:16] um the time I want the first video to

[05:19] start fading out so here in this case I

[05:22] have a video that is more or less 10

[05:24] seconds long so I can put it in the

[05:26] middle five seconds

[05:28] and in the end I can put the name of the

[05:31] parameters so the name of the parameters

[05:33] uh the code sorry the collect parameters

[05:37] leave x264

[05:41] preset

[05:43] slow minus CR f

[05:47] at 15 so not so so a fairly high quality

[05:52] but not the super high quality in order

[05:54] to keep this phase

[05:56] the space contained and

[06:01] then

[06:03] we can put the name of the files in this

[06:05] case we can say Singapore

[06:09] xfader.np4 and let's see what happens

[06:14] in this case we have an error what is

[06:16] the the error here it says the first

[06:19] input linked main time base do not match

[06:22] the corresponding second input link X3

[06:25] so we have the time base so the let's

[06:28] say the internal clock with which we are

[06:30] ticking the advancement from one frame

[06:33] and the other that is uh not set

[06:36] correctly here this is faster than this

[06:40] this has nothing to do with the frame

[06:43] rate is only the advancement in terms of

[06:46] time stamp that is present within the

[06:48] file so what it would like to do is uh

[06:50] to set the time base to the

[06:54] um

[06:54] photo audio video time based it usually

[06:57] is calculated with the frame rate if I'm

[07:00] not wrong so this is the what I found

[07:02] out so if you um

[07:04] correct me but this is a calculated

[07:07] based

[07:09] audio and the video time based

[07:13] and so what we can do is to apply this

[07:17] before the Crossfade so what can we do

[07:20] we are still or we can use the previous

[07:23] command so we have the

[07:26] Singapore one Singapore two files this

[07:30] uh still accepts the two inputs so here

[07:34] let's put a bit of space then later we

[07:36] are packing everything back

[07:39] so what shall we do with the first input

[07:41] with the first input we said that we

[07:43] want to set the time base in order to

[07:46] put it to the audio video time based so

[07:49] set TB

[07:51] equals to a b t b

[07:54] so capital letters and then we're going

[07:57] to put the set PTS

[08:01] to the B the beginning of the uh

[08:06] as you can do so start PTS

[08:09] in this case we need to give an output

[08:13] name so in this case it can be out

[08:18] zero

[08:19] the output file zero and then we got to

[08:23] put the uh

[08:26] semicolon the semicolon separates the

[08:29] different filters so we have a first

[08:33] um

[08:34] set TB

[08:36] and the set BTS then we got this output

[08:41] there is a

[08:43] by semicolon to the second output and in

[08:46] the set sorry the second input file

[08:47] second input file needs the same thing

[08:50] so in this case we have the set TV

[08:58] set PTS

[09:00] equal to the pts and start BTS

[09:06] and in this case we're going to call

[09:08] this out one so we have the output uh so

[09:13] we got just to sum up

[09:16] first input that we call the zero file

[09:19] zero that passes through the set time

[09:23] base set BTS and it exit as out zero

[09:29] then we got the second that is called

[09:32] one it's a right here second one then

[09:36] it's going it's going through the same

[09:38] thing so we've got the the set TB set

[09:41] time base and the set pts and it's

[09:45] called the out one so we got out zero

[09:48] and out one after we have the out zero

[09:50] and the out one we're going to merge

[09:52] them together in order to obtain a

[09:53] single output in this case there's a no

[09:57] need to call the output

[09:59] um

[10:01] because

[10:02] it's like a sink we put the output

[10:06] unnamed it and here

[10:09] we got to put the first input and the

[10:14] first a second in sorry the first output

[10:17] of the previous

[10:20] pipe and the output of the second pipe

[10:24] coming from the second file so here we

[10:26] got to put out zero

[10:28] and uh out

[10:33] Crossfade transition and blah blah with

[10:36] everything that we said before and let's

[10:38] see if now it's working

[10:40] it asks if you want to overwrite the

[10:42] file and we click yes

[10:44] and it seems the thing is working cool

[10:48] so we wait a bit of time until this is

[10:51] finished

[10:54] and uh you will see the speed

[10:57] yeah it is more or less recorded no

[10:59] video and now it's time to play uh

[11:03] Beautiful video so we've got the

[11:05] Singapore X fade

[11:08] and here we go with the first input

[11:13] there is a transition from the first

[11:16] video to the second

[11:17] cool so I got to say that this thing

[11:21] about the set time base is a bit uh

[11:24] say not the forecast it's something that

[11:28] I didn't know so

[11:30] um

[11:31] most normally this is a thing that you

[11:34] shouldn't do but the

[11:37] the Crossfade is the necessary one but

[11:40] so we saw a slightly more complex filter

[11:43] complex so-called filter graph

[11:46] that

[11:47] um

[11:50] fine

[11:51] that's fine feel the complex

[11:55] okay we have our Crossfade video we just

[11:58] created and now we are going to use it

[12:01] to make a video with the rgmb together

[12:06] all in a single video with a tiling so

[12:09] original video red Channel green Channel

[12:11] blue Channel

[12:14] and we use a filtered complex so to

[12:16] understand that how the filter complex

[12:19] are working so we are using FM pack and

[12:23] seeing Apple X fade.mp4 is our input

[12:26] file here we can repeat the input

[12:30] multiple times but we use the split

[12:32] filter which is a bit more clever and so

[12:36] it makes us understand better how the

[12:38] filter graph is working so we have the

[12:42] inputs and the outputs explained better

[12:45] with a complex field you're going to use

[12:46] so then we're going to use the filter

[12:49] complex

[12:52] we use the accent go to the new line the

[12:55] triangle accent

[12:58] and uh the next line would be the first

[13:02] line of the filter so here we put the

[13:05] quote and our input is the file so we

[13:11] have the zero then we put a split the

[13:15] split requires the outputs

[13:18] parameter the outputs parameter we have

[13:21] three channels so we have a one two and

[13:24] three in this case uh how can we call

[13:26] them we call them

[13:28] um let's say uh V1

[13:32] V2 and V3

[13:35] so we have a

[13:39] zero there is an input then B1 V2 V3

[13:42] where that will be the input for the

[13:45] next filters and what are the next

[13:47] filters the next filters will be

[13:51] so the filter is a loot RGB

[13:57] if this stands for lookup table RGB in

[14:01] this case we can make correspondences

[14:03] between inputs and outputs uh

[14:06] different channels in the RGB color

[14:08] space and uh

[14:10] so root RGB

[14:13] this is a red channel so we got to put

[14:15] the green at zero

[14:18] and the blue at zero

[14:20] and this will be our red Channel

[14:23] and

[14:24] colon

[14:26] Uline

[14:29] we have a steal the loot RGB where we

[14:33] put in this case the red is zero and the

[14:36] blue at zero since we want the the green

[14:39] Channel

[14:43] RG

[14:47] okay and the third line will be V3

[14:53] loot RGB equals to a red of zero and a

[15:00] green of zero since we want the blue

[15:03] Channel

[15:05] and now we separate the three rgmp

[15:10] then now we want to put on the top line

[15:14] the input and the red channel so we put

[15:17] a h stack horizontal stack of these two

[15:19] inputs and the same thing will be done

[15:22] at the bottom where the the green and

[15:24] the blue channel so here the next will

[15:28] be the input video Zero and a red

[15:34] Channel

[15:36] this case it is on each stack

[15:39] equal to inputs

[15:42] so this is the parameter that we need to

[15:45] accept the H stack accepts the inputs

[15:48] equals to two since we have the uh zero

[15:53] and Dr and

[15:55] we name it top

[15:58] oh

[16:01] okay and we do the same thing with the G

[16:06] and the B

[16:09] right here there is on each stack

[16:12] with two inputs

[16:17] and this will be our bottom

[16:21] okay

[16:23] and now the final filter uh the final

[16:26] one will be a vertical stack so we put

[16:28] the top and the bottom together

[16:31] and so we have the top and we have the

[16:37] bottom

[16:38] with this stack equals to

[16:43] inputs to the same as each stack to be

[16:46] in the v stack go to two here we can

[16:50] select if we wanted to give an output to

[16:54] the output a name or not so here we can

[16:56] for example put out

[16:59] is a output but you got to keep in mind

[17:02] that here we got to put a map

[17:04] out

[17:06] in this way if you want you can put it

[17:09] in this case just to avoid since the

[17:11] notation is quite long and putting

[17:14] something like this is a bit more

[17:16] um

[17:17] shorter and won't stand back we

[17:19] understand better what it's doing since

[17:21] the output is one so it's okay

[17:25] now it's time to put our options so

[17:28] there is the lead backs 264

[17:31] preset

[17:33] let's put the slow

[17:37] and as a constant rate Factor we put 18

[17:42] so a fairly good quality but not the

[17:44] maximum just to contain the size of the

[17:46] file and finally we give a name

[17:49] [Music]

[17:50] to do it so we can it can be Singapore

[17:53] it's fader RGB

[17:57] as format we can put MKV let's see

[18:01] how it can be

[18:04] good uh I think everything is okay so to

[18:07] sum up we have the splitting three

[18:09] channels then we take the red green and

[18:13] blue of the three Channel we just output

[18:15] from the split then we have the

[18:19] horizontal Stacks we put the two videos

[18:22] together respectively the input and the

[18:25] red channels

[18:27] um then we have the green and the blue

[18:29] together and then we

[18:31] see if it is working and it's working

[18:34] now we wait for the output to be

[18:36] generated uh now maybe it is a

[18:41] bit slow but the video is not that long

[18:44] and we wait until the end of the

[18:46] processing

[18:51] [Music]

[19:01] okay now the processing has finished and

[19:04] now it's time to play the video so I

[19:06] think that the video will be too big for

[19:09] the screen so

[19:11] um I'm trying to open it with the oclc

[19:18] and here we go we have the input the red

[19:22] channel the green Channel and the blue

[19:25] as we wanted again if you would like to

[19:28] take a look about how the complex

[19:30] filters work and to build your

[19:32] photograph so again we make a sum up

[19:36] so to set your filter complex it should

[19:40] be set with a filter from Flex

[19:45] and then here we put the input

[19:50] then the name of the filter

[19:53] with its parameter iron one

[19:57] uh equals to Value one then the second

[20:01] value will be in the

[20:04] uh

[20:05] a colon then W2

[20:09] Etc and here we put our output

[20:14] if we generate from a single input

[20:17] multiple outputs like the split filter

[20:19] we put the out one and out to

[20:25] so here we have all then

[20:29] with a semicolon we decided what to put

[20:31] in the next filter and here we put

[20:36] out one

[20:39] and

[20:41] I'll

[20:42] then we put the name of the filter and

[20:44] so on and so forth so the fundamental

[20:46] thing is

[20:48] input here then output output or outputs

[20:54] after the filter itself and we go on to

[20:58] have a better picture about the filters

[21:02] that we have here in contact we can go

[21:05] with FM pack minus h

[21:09] feel this no I'm sorry

[21:12] if I met if I'm packed minus filters

[21:16] will show the list of all the filters

[21:18] that we have available if you want to

[21:21] look for the filter of choice in Windows

[21:24] you can use with a pipe fine Str

[21:28] ucture let's find Str

[21:32] Ike this

[21:33] um let's say split

[21:37] it will find the

[21:40] dirt is up to you so here we have a

[21:42] single video with a multiple video

[21:45] outputs and here we have also

[21:49] things regarding the audio so same thing

[21:52] that we did with the video we can do it

[21:54] with the audio so there's a cross fader

[21:57] for the audio as well

[21:59] it's called I think a cross fade check

[22:05] across fader

[22:08] um we have a Crossfade of two you've got

[22:10] the audio streams and here we got the

[22:12] the X faded ah we also have a x-fit

[22:15] opens

[22:19] parameters if I want to check we have

[22:22] the final meters here we have X fader so

[22:26] the thing is a FM pack minus h

[22:31] ilton equals to Theta

[22:36] and we can take a look at the

[22:44] type okay it's a transition

[22:51] who

[22:53] okay so we saw something about the

[22:55] filter complex

[22:57] um I hope you enjoyed this one of the

[22:59] most complicated things you can do with

[23:02] fmpeg actually but it will help you a

[23:05] lot in case you want to automate your uh

[23:09] video generation process so if you would

[23:11] like to make crossfades you want to make

[23:15] fancy facts to your video

[23:17] substantially sometimes you want to make

[23:20] a slideshow of your vacation video

[23:23] essentially you're doing something like

[23:24] this or either you concatenate or you

[23:27] Crossfade from one video to the other

[23:29] there might be something else to say

[23:31] about the cross fading if you want to do

[23:33] it in a

[23:35] good way so maybe it's a bit

[23:40] uncomfortable to have the offset from

[23:44] the beginning of the first video

[23:47] um but knowing if you want to make a

[23:49] concatenation with a Crossfade or

[23:51] something but anyway this is a powerful

[23:54] tool you can look yourself on how to do

[23:56] it so it's like an overlay

[24:01] anyway thank you very much if you reach

[24:02] up to this point I hope you liked it if

[24:05] you liked it like share and subscribe

[24:06] and hit the notification Bell to be

[24:09] always notified with the new video I'll

[24:12] be releasing and um

[24:15] I remember you to subscribe and follow

[24:17] me on the social media uh I maintaining

[24:20] and trying to maintain at least so

[24:22] Instagram and Tick Tock uh for new

[24:25] content

[24:27] here and uh I hope to see you next time

[24:30] with new content relative to multimedia

[24:33] and much much more thank you very much

[24:36] and ciao ciao

[24:43] [Music]

[24:51] foreign