VideoTek (Avisynth and MPV Waveform Monitor)

[FranceBB]

Hi,
I love the Tektronix Videoscope I have at work, but I can't use it at home, that's why I tried to reproduce a similar output using Avisynth built-in videoscope (Histogram).
Please welcome VideoTek, a Tektronix-like waveform monitor for Avisynth and MPV.

VideoTek() (Avisynth Version) - VideoTek (MPV Version)



picture (top left), luma (top right)
chroma (bottom left), audio channels and Lissajou (bottom right)


Avisynth Usage:

just call the function: VideoTek() in your script


MPV Usage:

Press "Y" to turn on the waveform monitor while you're playing back a video.


Please note that due to the limitation of LAV Filter Complex, the MPV version is slightly different from the Avisynth Version:




Changelog:

v0.5 Crappy original version by FranceBB
v0.6 Commit by magiblot
v0.7 Major code cleanup by StainlessS
v0.8 Added 6ch display with the graph of the first two by Stainless
v0.9 Fixed a bug occurring with 1ch in the Lissajou graph by Stainless
v1.0 Added yv16 and yv24 planar 8bit support by FranceBB
v1.1 Introduced High Bit Depth Support, Frame Properties, Timecode and Luma Values in Volts by FranceBB
v1.2 Added support to all color spaces and code cleanup by StainlessS
v1.3 Added MPV Support by Algie
v1.4 Added frame properties support by FranceBB
v1.5 Improved small monitor image conversion and fixed a bug that would sometimes report invalid info at the bottom on the left by FranceBB
v1.6 Added Fractional Framerates Support.
v1.7 Implemented the new VideoTek(Mode="PQ") with the new 0.406V 58% Reference White marker
v1.8 Implemented new "Type" parameter to allow nits display instead of volts VideoTek(Mode="HLG", Type="nits"), but by default Type="volts".
v1.9 Introduced Type="nits" support for the recently introduced PQ Mode to display the 0, 203, 1000, 5000 and 10000 nits markers and Detailed=true to see all the 1k nits increments.

Cheers.

[lansing]

To me the real deal of Tektronix's hardware is the double diamond vectorscope, it makes color matching looks so easy. But I think they patented it.
https://www.youtube.com/watch?v=8NTLqybQQSU

[FranceBB]

@lansing... yep, that's really useful, but I mainly work on post-production, so this is how I set it:

Tektronix - Avisynth

Another example: https://i.imgur.com/EF3VWbu.png

[lansing]

Taking more look into the screenshot, the proportion of the graphs should be independent from the video window, they shouldn't be scaled 16:9 along with the video. And there needs to be boxes in the vectorscope to indicate the legal saturation range for standard definition and high definition. Also the wave form needs to have a some kind of indication of brightness, such as percentage.

I have a video only clip and it just return an error because I don't have an audio.

[FranceBB]

I updated it to a newer version. It's supposed to automatically detect the audio channels number and make a graph accordingly, but... it doesn't work... I think I broke my own code xD

Do you have any idea why it doesn't show the audio graph anymore?

[magiblot]

Well, you store things into the my_audio_levels_overlayX variables and then leave them unused.

[FranceBB]

That's the thing. I'd like to use Overlay(audio_levelsx) according to how many channels are detected, but I don't know how to do it.

[magiblot]

No problem

Try with the following changes:

source_audio_channels = clp.AudioChannels() #Does this return 0 if there is no audio?

myaudio = (source_audio_channels==0) ? GetChannels(myfakeaudio, 1, 2) :
\ (source_audio_channels==1) ? GetChannels(clean, 1) :
\ (source_audio_channels==2) ? GetChannels(clean, 1, 2) :
\ (source_audio_channels==3) ? GetChannels(clean, 1, 2, 3) :
\ (source_audio_channels==4) ? GetChannels(clean, 1, 2, 3, 4) :
\ (source_audio_channels==5) ? GetChannels(clean, 1, 2, 3, 4, 5) :
\ (source_audio_channels==6) ? GetChannels(clean, 1, 2, 3, 4, 5, 6) :
\ (source_audio_channels==7) ? GetChannels(clean, 1, 2, 3, 4, 5, 6, 7) :
\ (source_audio_channels==8) ? GetChannels(clean, 1, 2, 3, 4, 5, 6, 7, 8) :
\ NOP()

for_audio_levels = AudioDub(myblack, myaudio)

audio_levels = for_audio_levels.Histogram("audiolevels").Crop(0, 0, -736, -0).BilinearResize(340, 420)

my_audio_levels_overlay = Overlay(audio_levels, x=700, y=450, opacity=1.0, mode="blend", greymask=true, ignore_conditional=false, pc_range=false)


final=Overlay(my_audio_levels_overlay) #you will probably want to add additional parameters

There could be syntax errors, I edited this from mobile.

[StainlessS]

Where is 'mylissajou' supposed to come from ? (what histogram or whatever plug).

EDIT: Histogram(mode="StereoOverlay") ???

[StainlessS]

Looks about right, might need [EDIT: WILL NEED] some adjustment to dimension, overlay coords.
Have used Histogram(mode="StereoOverlay") for the mylissajou thing, hope that its correct.

#simple Videoscope by FranceBB:- https://forum.doom9.org/showthread.php?t=175249
#it resembles Tektronix WVR 7020
#Accepts 8bit clip only. It works internally in 4:2:0 (yv12)
#Accepts up to 8 audio channels

Function VideoTek(clip clp) {
    clp
    nChannels=(HasAudio)?AudioChannels:0                        # ssS, HasAudio = audio_samples_per_second!=0
    #Assert block
    Assert(IsYV12,"VideoTek: source must be yv12 (YUV 4:2:0 8bit)")
#   Assert(IsFrameBased, "VideoTek: source must be progressive") # ssS, IsFrameBased means SeparateFields NOT used, does not indicate Progressive
    Assert(nChannels<9,"VideoTek: sources with more than 8 audio channels are not supported")

    SilentStereo=Last.BlankClip(width=1480, height=920,audio_rate=48000,channels=2,sample_type="Float")
    myBlack= (nChannels>0)  ? SilentStereo.AudioDubEx(Last) : SilentStereo.Killaudio
    Stereo = (nChannels>=2) ? Last.GetChannels(1,2) : (nChannels==1) ? MergeChannels(GetChannels(1),GetChannels(1)) : SilentStereo
    StereoClip = Last.BlankClip(width=460,height=460).AudioDub(Stereo)  # Histogram(mode="StereoOverlay") requires 2 channel only clip.

    clean1=PointResize(704, 396)
    Picture    = clean1
    luma       = clean1.TurnRight.Histogram(mode="classic").TurnLeft.Crop(0,0,0,256).BilinearResize(704, 396)
    chroma     = clean1.Histogram("color2").Crop(704, 0, 256,256).BilinearResize(480,480)
    Audio      = myBlack.Crop(0,0,340,420).Histogram(Mode="audiolevels")
    mylissajou = StereoClip.Histogram(mode="StereoOverlay")  # Dont know where to get lissajou from, so improvised

    Result = MyBlack
    Result = Result.Overlay(Picture,    x=0, y=0)        # ssS, Removed all of the Optional Args assigned with Defaults
    Result = Result.Overlay(luma,       x=710, y=0)
    Result = Result.Overlay(chroma,     x=100, y=430)
    Result = Result.Overlay(Audio,      x=700, y=450)
    Result = Result.Overlay(mylissajou, x=1000,y=450)
    return Result
}

EDIT: BugFixed : EDIT: Same bug fixed AGAIN !!!

Untested with more than 2 channels. [EDIT: or with single channel, tested on 0 and 2 channels only]

ColorBars(Pixel_Type="YV12")
VideoTek
Return Last.Spline64Resize(854,480)

EDIT: Below bottom right image is more like what I remember as lissajou figure [EDIT: perhaps samplerate just too low]


EDIT:

Stereo = (nChannels>=2) ? Last.GetChannels(1,2) : SilentStereo

Not sure how to process/mix multiple channels into stereo, above, just selected Left and Right.

[FranceBB]

Thank you magiblot and StainlessS

@StainlessS... yes, it's Histogram(mode="StereoOverlay"). The idea is to pick both Left and Right channel when stereo and pick just Front Left and Front Right when is Dolby.

It might not be as complex and detailed as the official waveform provided by Tektronix, but it kinda reminds its layout and it uses the internal filters already available in Avisynth, so I thought it was kinda cool to have a similar layout. ^_^

Thank you both, once again, this is exactly what I wanted. ^_^

[StainlessS]

bugFix,

was

    Stereo = (nChannels>=2) ? Last.GetChannels(1,2) : (nChannels==2) ? MergeChannels(GetChannels(1),GetChannels(1)) : SilentStereo

should have been

    Stereo = (nChannels>=2) ? Last.GetChannels(1,2) : (nChannels==1) ? MergeChannels(GetChannels(1),GetChannels(1)) : SilentStereo

I'll leave it to you to perfect it.
Fixed in prev post.

EDIT: Your original Big Buck Bunny Link (downsized to 854,480)


And current script(approx same frame, needs some coords fixed a bit)



EDIT:

The idea is to pick both Left and Right channel when stereo and pick just Front Left and Front Right when is Dolby.

From Avs Docs

Every file format has a different internal channel ordering. The following table gives this internal ordering for some formats (useful for plugin writers), but it is a decoder task supply to AviSynth the expected channel order (like WAV), if you use decoders like NicAudio/BassAudio or ffdshow/AC3_filter you don't need to worry about this:

reference:      channel 1:     channel 2:     channel 3:     channel 4:     channel 5:     channel 6:
5.1 WAV         front left     front right    front center      LFE         rear left      rear right

So I take it that channels 1 and 2 should be correct whether DOLBY or not (as we are dealing with WAV order).

[StainlessS]

Bit more fiddling about.

Function VideoTek(clip clp) {
    clp
    nChannels=(HasAudio)?AudioChannels:0                        # ssS, HasAudio = audio_samples_per_second!=0
    Assert(IsYV12,"VideoTek: source must be yv12 (YUV 4:2:0 8bit)")
    Assert(nChannels<9,"VideoTek: sources with more than 8 audio channels are not supported")
    CSZ = 480                 # Chroma Size box (W & H)
    A_W = 320                 # Audio Box Width
    A_H = 460                 # Audio Box Height (Audio & Lissajoua)
    L_W = (704+704+6)-700-A_W # Lissajous Width
    SilentStereo=Last.BlankClip(width=1480, height=920,audio_rate=48000,channels=2,sample_type="Float")
    myBlack= (nChannels>0)  ? SilentStereo.AudioDubEx(Last) : SilentStereo.Killaudio
    Stereo = (nChannels>=2) ? Last.GetChannels(1,2) : (nChannels==1) ? MergeChannels(GetChannels(1),GetChannels(1)) : SilentStereo
    StereoClip = Last.BlankClip(width=16,height=16).AudioDub(Stereo)  # Histogram(mode="StereoOverlay") requires 2 channel only clip.
    Picture   = BicubicResize(704, 396)
    luma      = TurnRight.Histogram(mode="classic").TurnLeft.Crop(0,0,0,256).BiCubicResize(704, 396)
    chroma    = Histogram("color2").Crop(clp.Width, 0, 0,256).BilinearResize(CSZ,CSZ)
    Audio     = myBlack.Crop(0,0,A_W,A_H).Histogram(Mode="audiolevels")
    lissajous = StereoClip.Histogram(mode="StereoOverlay").BicubicResize(L_W,A_H)
    MyBlack
    Overlay(Picture,   x=0,                   y=0)
    Overlay(luma,      x=704+6,               y=0)
    Overlay(chroma,    x=(704-chroma.Width)/2,y=396 + (920-396-chroma.height)/2)
    Overlay(Audio,     x=704-4,               y=396 + (920-396-audio.height) /2)
    Overlay(lissajous, x=700+Audio.Width,     y=396 + (920-396-lissajous.height)  /2)
    return Last
}

Can mess with this lot to auto center stuff, when happy then might want to hard code numbers.

    CSZ = 480                 # Chroma Size box (W & H)
    A_W = 320                 # Audio Box Width
    A_H = 460                 # Audio Box Height (Audio & Lissajoua)
    L_W = (704+704+6)-700-A_W # Lissajous Width

Have also converted from PointResized source to input source for histogram stuff, and used Bicubic instead where resize necessary,
if too slow, then change BiCubics to BiLinear.

EDIT: Added Levels() on Lissajous trying to make more visible, only minor improvement (before downsize).

Here, downsized,


EDIT:
In Steinberg, I think that the Lissajous thing is known as Phase (or Phase Scope), when spreads out horizontally, is broader stereo,
if dot or vertical line, then is mono, height being amplitude (I guess FranceBB, already knows this).
The Phase Scope, shows info from only first two channels (Left & Right, it is stereo specific, perhaps eg 5.1 should be downmixed somehow).

EDIT: Moved Levels() before BiCubicResize() for Lissajous.
(Levels improvement is only on the cross hair thing, is a bit more visible on FranceBB linked image than in this script)

EDIT:
Changed Levels 126-4 and 126+4, to 126-1, 126+1. (126 being TV levels mid point)

lissajous = StereoClip.Histogram(mode="StereoOverlay").Levels(126-1,1.0,126+1,16,235,coring=false).BicubicResize(L_W,A_H)

EDIT: Damn, below bug crept back in again, fixed in blue [was (nChannels==2)].

Stereo = (nChannels>=2) ? Last.GetChannels(1,2) : (nChannels==1) ? MergeChannels(GetChannels(1),GetChannels(1)) : SilentStereo

EDIT: Levels() removed from Lissajous thing, it emphasized the cross hairs stuff, but messed with the dust cloud a bit.

[StainlessS]

FrancBB, I see your there,
NOTE, the bug that crept back in is also present in your 1st post, must have edited on-line and forgot to make same bug fix in my script
and then made mod with unfixed script.
(2nd last edit in above post).

[FranceBB]

when spreads out horizontally, is broader stereo,
if dot or vertical line, then is mono, height being amplitude (I guess FranceBB, already knows this).

Yes, I use it on a daily basis at work to check whether there's difference between channels in a stereo pair or not.

NOTE, the bug that crept back in is also present in your 1st post

Oh, I see...

So I take it that channels 1 and 2 should be correct whether DOLBY or not (as we are dealing with WAV order).

Yes, it's always gonna follow the wav order 'cause Avisynth internally works with this order, so, if someone indexes a video file that has an AC3 Dolby audio track with an indexer, it's up to the indexer to return a WAV audio order to Avisynth.



Anyway, thank you for modifying the code. I owe you a beer.
Perhaps, one day, when there'll be perfect 3D scanners... xD

[reisub]

FranceBB, thanks for the script. At work, for quality control, I use mainly Aurora Tektronix. But I was pleasantly surprised by the possibilities of this script.

[FranceBB]

Version 1.1 Released:
- Introduced High Bit Depth Support (***)
- Introduced Frame Properties and Timecode
- Introduced Luma Values in Volts


*** I actually cheated as I'm doing things in 8bit planar, however it works as bit-shifting is done in a way that preserves fullrange/limitedrange values


In the bottom left, you're now gonna see the resolution, framerate, colorspace and bit depth of the clip you indexed and on the bottom right you're gonna see the timecode which scales according to the framerate, just like you would in the Tektronix waveform monitor (or in a Sony VTR).

Originally Posted by reisub

FranceBB, thanks for the script. At work, for quality control, I use mainly Aurora Tektronix. But I was pleasantly surprised by the possibilities of this script.

I'm glad you like it. I wanted to bring those waveform monitors to Avisynth so that I could check the files I encode before sending them to the QC guys without having to be in the office with my workstation connected to a piece of hardware via SDI.

[StainlessS]

FBB,
Some Comments flagged by ssS:

Function VideoTek(clip clp) {
    clp
    nChannels=(HasAudio)?AudioChannels:0                        # HasAudio = audio_samples_per_second!=0

    #Errors checking
    Assert(IsPlanar && IsYUV,   "VideoTek: source must be planar YUV")       # ssS: IsPlanar simpler than IsYV12 etc
    Assert(nChannels<9,"VideoTek: sources with more than 8 audio channels are not supported")

    #Getting info
    #Resolution
    x=clp.Width()
    y=clp.Height()
    #Framerate
    fps=clp.FrameRate()                                         # ssS: Not FrameRateNumerator (check Src with eg Assumefps(30000,1001)
    #ColorSpace
    colorspace=clp.PixelType()

    #Taking care of High bit Depth
    Bpc = clp.BitsPerComponent
    clp.ConvertBits(bits=8)                                     # ssS: If already 8 bits then ConvertBits should do nothing anyway (return original)


# ...

See SendSpace below sig for Avs (not avsi, remove my comments and test code, and rename to avsi).

Appended testcode after your script, in the avs

AviSource("d:\Parade.avi")

Assumefps(30000,1001)           # Test FrameRate

# UnComment ONLY ONE of below

#ConvertToRGB32                                         # RGB32
#ConvertToRGB24                                         # RGB24
#ConvertToYUY2                                          # YUY2
#ConvertToYV12                                          # YV12
# v2.60 +
#ConvertToYV411                                         # YV411
#ConvertToYV16                                          # YV16
#ConvertToYV24                                          # YV24
#ConvertToY8                                            # Y8
# Avs+
#ConvertToRGB48                                         # RGB48     Intereleaved RGB  16 bit
#ConvertToRGB64                                         # RGB64     Intereleaved RGBA 16 bit

# ... + Loads more

# 32 Bit Float
#ConvertToRGB48.ConvertToPlanarRGB.ConvertBits(32)      # RGBPS
#ConvertToRGB64.ConvertToPlanarRGBA.ConvertBits(32)     # RGBAPS
#ConvertToY8.ConvertBits(32)                            # Y32
#ConvertToYV12.ConvertBits(32)                          # YUV420PS
#ConvertToYV16.ConvertBits(32)                          # YUV422PS
#ConvertToYV24.ConvertBits(32)                          # YUV444PS
#ConvertToYV12.ConvertBits(32).AddAlphaPlane            # YUVA420PS
#ConvertToYV16.ConvertBits(32).AddAlphaPlane            # YUVA422PS
ConvertToYV24.ConvertBits(32).AddAlphaPlane            # YUVA444PS
###################

Videotek

EDIT: Fix your avsi script and I'll remove this post.

EDIT:
More mods, posted on SendSpace. [also, removed references to clp after initializing Last].

Function VideoTek(clip clp) {
    clp
    nChannels=(HasAudio)?AudioChannels:0                        # HasAudio = audio_samples_per_second!=0

    #Errors checking                                            # ssS: Might as well do at beginning (after errchk, known Planar YUV & < 9 channels)
#    Assert(IsPlanar && IsYUV,   "VideoTek: source must be planar YUV")       # ssS: IsPlanar simpler than IsYV12 etc
    Assert(nChannels<9,"VideoTek: sources with more than 8 audio channels are not supported")

    #Getting info
    #Resolution
    x=Width()
    y=Height()
    #Framerate
    fps=FrameRate()                                             # ssS: Not FrameRateNumerator (check Src with eg Assumefps(30000,1001)
    #ColorSpace
    colorspace=PixelType()

    #Taking care of High bit Depth
    Bpc = BitsPerComponent


 #   ConvertBits(bits=8)                                     # ssS: If already 8 bits then ConvertBits should do nothing anyway (return original)
    ConvertToYUV444.ConvertBits(bits=8)                          # ssS: Convert To YV444, 8 bit, support all colorspace input

 # ...

Supports ALL input colorspace, not just Planar YUV. [Only tested with YUY2, RGB64, RGBPS + a few more, but all should work]

EDIT: Fixed Error in either Avs+ or VD2.
Bug report here in devs forum:- https://forum.doom9.org/showthread.p...58#post1927858

fix [Updated script on SendSpace]

Function VideoTek(clip clp) {
    clp
    nChannels=(HasAudio)?AudioChannels:0                        # HasAudio = audio_samples_per_second!=0

    #Errors checking
    Assert(nChannels<9,"VideoTek: sources with more than 8 audio channels are not supported")

    #Getting info
    #Resolution
    x=Width()
    y=Height()
    #Framerate
    fps=FrameRate()
    #ColorSpace
    colorspace=PixelType()

    #Taking care of High bit Depth
    Bpc = BitsPerComponent

    ConvertToYUV444
    (HasAlpha) ? RemoveAlphaPlane : NOP          # ssS: Avoid eg YUVA444 error in VirtualDub2 (OK in PotPlayer but grey screen in VD2 if YUVA)
    ConvertBits(bits=8)                          # ssS: Convert To YUV444, 8 bit, NO ALPHA, support all colorspace input

 # ...

[FranceBB]

Woah, thank you for that!
I just tested it, it works like a charm.
I've just updated the code on GitHub.

Version 1.2 released.
Updated Changelog:
v1.2 Added support to all planar color spaces and code cleanup by StainlessS

p.s if you have a GitHub account, I'll give you full access to the repository
p.p.s you know why I used millivolt (or rather Volts) as a measurement instead of putting 16-235 in the luma monitor? 'cause Volt levels are the same for each signal, no matter the bit depth, while bit values (16-235) change with bit depth (like 10bit, 64-940 etc) and I didn't want to write a conditional based on that xD Still, the Tektronix engineers did the same and showed Volts as scale, so... I guess it's fine. xD

[StainlessS]

v1.2 Added support to all planar color spaces ...

Should be
"v1.2 Added support for all input color spaces ..."

I will not remove my prev post, as it gives some background to bugrep in devs forum.

EDIT: Nope, no GitHub account just yet, one day maybe.

Temp VideoTek.avs removed from SendSpace.