add motion blur on 50P video?

[marcorocchini]

Hi,

<-- a cat here # please be patient. Thanks.

Please download my test footage, 1 file is 25P and 1 is 50P

https://www.swisstransfer.com/d/0e287ac9-0b5a-4d36-a2da-c9d71fee8812

I wonder: please consider the 50Pm, after edit it at 50P, and export to 50P can I add a little motion blur to make it appear less "live" and a little more cinematic?

[manono]

Motion Blur can be added in AviSynth using QTGMC:

"Motion-blur can also be applied to double-rate output, but the effect is less noticeable."
"QTGMC( Preset="Slower", FPSDivisor=1, ShutterBlur=1, ShutterAngleSrc=0, ShutterAngleOut=720 ) # Blur some high shutter-speed footage, keep double rate"

avisynth.nl/index.php/QTGMC

[marcorocchini]

I forgot to say that my target ("unfortunately") is the 1080i25 format

so in consideration of this, what should the avisynth script for the 50P.MXF?

[pandy]

you can use ffmpeg and filter 'tmix' i would also try to bump framerate (with for example 'minterpolate' filter), add some tmix "motion blur" so overall motion will be perceived smoother and later converted to interlace.

[marcorocchini]

my target is to get a "cinematic" aspect from my 50P video, What can be the script?

[marcorocchini]

I tryed to set the camera into 25P but video appear with too motion blur. If I set the camera to 50P video appear too "live".

My target is delivery on 50i and interlaced video does match very well the 25p or 50p footage, so my so my hypothetical idea would be to shoot and edit and export in 50P, then eventually convert to 50i with a little cinematic motion blur, if possible

[Sharc]

In Avisynth try this to convert your 50p source to 25i (sometimes called 50i for 50 fields per second) and see how it looks on your TV:

<your source filter for your 50P video here>
SeparateFields()
SelectEvery(4,0,3)
Weave() #interlaced video with 50 fields per second

[marcorocchini]

In Avisynth try this to convert your 50p source to 25i (sometimes called 50i for 50 fields per second) and see how it looks on your TV:

<your source filter for your 50P video here>
SeparateFields()
SelectEvery(4,0,3)
Weave() #interlaced video with 50 fields per second

yes this works but it does not add a "cinematic" look, it appears a little too "fluid" and too "live" than a cinematic look.
If possible I try to get a litte motion blur from footage 50P that is too "real" than the 25P footage that appear with too many motion blur

[marcorocchini]

can I use QTGMC

[jagabo]

QTGMC, http://avisynth.nl/index.php/QTGMC#Shutter_Speed_Motion_Blur_.2F_Frame_Rate
Doesn't work well with your video.

Try RIFE() to change the fps to 400 then Successive Merge() and/or SelectEven() to reduce the frame rate.

function SmoothFPS_RIFE(clip source, int num, int den)
{ 
    source
    z_ConvertFormat(pixel_type="RGBPS", colorspace_op="709:709:709:l=>rgb:709:709:f")
    Rife(gpu_thread=1, model=9, gpu_id=0, fps_num=num, fps_den=den, sc=false, sc_threshold=0.10)
    z_ConvertFormat(pixel_type=source.pixeltype, colorspace_op="rgb:709:709:f=>709:709:709:l")
}

LWLibavVideoSource("DJI_20250216183522_0078_D.MP4") 
SmoothFPS_RIFE(400000, 1000) # 400 fps
Merge(SelectEven(),SelectOdd()) # 200fps 
Merge(SelectEven(),SelectOdd()) # 100 fps
Merge(SelectEven(),SelectOdd()) # 50 fps
SelectEven() # or Merge(SelectEven(),SelectOdd() again for more blur, 25 fps

That gives you 25p with motion blur and the jerky film look.

Of course, the correct way of doing this is to shoot at the frame rate you want and force the shutter speed to give the amount of motion blur you want. For example, 25 fps with 1/50 second shutter.

[marcorocchini]

oh wow

LoadPlugin("v:\x64\LSMASHSource.dll")




function SmoothFPS_RIFE(clip source, int num, int den)
{ 
    source
    z_ConvertFormat(pixel_type="RGBPS", colorspace_op="709:709:709:l=>rgb:709:709:f")
    Rife(gpu_thread=1, model=9, gpu_id=0, fps_num=num, fps_den=den, sc=false, sc_threshold=0.10)
    z_ConvertFormat(pixel_type=source.pixeltype, colorspace_op="rgb:709:709:f=>709:709:709:l")
}

LWLibavVideoSource("C:\Users\Administrator\Desktop\Nuova cartella (3)\DJI_20250216183449_0077_D.MP4")
SmoothFPS_RIFE(400000, 1000)
Merge(SelectEven(),SelectOdd()) # 200fps 
Merge(SelectEven(),SelectOdd()) # 100 fps
Merge(SelectEven(),SelectOdd()) # 50 fps
SelectEven() # or Merge(SelectEven(),SelectOdd() again for more blur, 25 fps

but I get


[Attachment 85639 - Click to enlarge]

what can it be?

[Sharc]

You need to put avsresize.dll into the Avisynth plugin folder.
http://avisynth.nl/index.php/Avsresize

[marcorocchini]

and also RIFE I think

I have installed
--> AviSynthPlus_3.7.3_20230715.exe

--> download RIFE-1.2.2.7z and placed RIFE.DLL in a specific folder

--> download avsresize_r21.7z and placed avsresize.dll in a specific folder

--> download models_from_0_to_52_and_sudo.7z and models_from_53_to_65.7z and decompress to the RIFE folder

this is the script:

LoadPlugin("v:\automazioneclip\AviSynth\LSMASH_AGO_2023\x64\LSMASHSource.dll")

LoadPlugin("V:\automazioneclip\AviSynth\RIFE\RIFE.dll")
LoadPlugin("V:\automazioneclip\AviSynth\RIFE\avsresize.dll")


function SmoothFPS_RIFE(clip source, int num, int den)
{ 
    source
    z_ConvertFormat(pixel_type="RGBPS", colorspace_op="709:709:709:l=>rgb:709:709:f")
    Rife(gpu_thread=1, model=9, gpu_id=0, fps_num=num, fps_den=den, sc=false, sc_threshold=0.10)
    z_ConvertFormat(pixel_type=source.pixeltype, colorspace_op="rgb:709:709:f=>709:709:709:l")
}

LWLibavVideoSource("C:\Users\Administrator\Desktop\Nuova cartella (3)\DJI_20250216183449_0077_D.MP4")
SmoothFPS_RIFE(400000, 1000)
Merge(SelectEven(),SelectOdd()) # 200fps 
Merge(SelectEven(),SelectOdd()) # 100 fps
Merge(SelectEven(),SelectOdd()) # 50 fps
SelectEven() # or Merge(SelectEven(),SelectOdd() again for more blur, 25 fps

but I get this error:


[Attachment 85645 - Click to enlarge]

[marcorocchini]

okok

I have locate the rife.dll and avsresize.dll as here


[Attachment 85646 - Click to enlarge]

and create a sub folder as here

[Attachment 85647 - Click to enlarge]

re-try to load this script:

LoadPlugin("v:\automazioneclip\AviSynth\LSMASH_AGO_2023\x64\LSMASHSource.dll")

LoadPlugin("V:\automazioneclip\AviSynth\RIFE\RIFE.dll")
LoadPlugin("V:\automazioneclip\AviSynth\RIFE\avsresize.dll")


function SmoothFPS_RIFE(clip source, int num, int den)
{ 
    source
    z_ConvertFormat(pixel_type="RGBPS", colorspace_op="709:709:709:l=>rgb:709:709:f")
    Rife(gpu_thread=1, model=9, gpu_id=0, fps_num=num, fps_den=den, sc=false, sc_threshold=0.10)
    z_ConvertFormat(pixel_type=source.pixeltype, colorspace_op="rgb:709:709:f=>709:709:709:l")
}

LWLibavVideoSource("C:\Users\Administrator\Desktop\Nuova cartella (3)\DJI_20250216183449_0077_D.MP4")
SmoothFPS_RIFE(400000, 1000)

Merge(SelectEven(),SelectOdd()) # 200fps 
Merge(SelectEven(),SelectOdd()) # 100 fps
Merge(SelectEven(),SelectOdd()) # 50 fps

SelectEven() # or Merge(SelectEven(),SelectOdd() again for more blur, 25 fps

but I get:


[Attachment 85648 - Click to enlarge]

what can it be?

[marcorocchini]

I try to use model=11 instead of model=9

don't know the differences but virtualdub open the avs

LoadPlugin("v:\automazioneclip\AviSynth\LSMASH_AGO_2023\x64\LSMASHSource.dll")

LoadPlugin("V:\automazioneclip\AviSynth\RIFE\RIFE.dll")
LoadPlugin("V:\automazioneclip\AviSynth\RIFE\avsresize.dll")


function SmoothFPS_RIFE(clip source, int num, int den)
{ 
    source
    z_ConvertFormat(pixel_type="RGBPS", colorspace_op="709:709:709:l=>rgb:709:709:f")
    Rife(gpu_thread=1, model=11, gpu_id=0, fps_num=num, fps_den=den, sc=false, sc_threshold=0.10)
    z_ConvertFormat(pixel_type=source.pixeltype, colorspace_op="rgb:709:709:f=>709:709:709:l")
}

LWLibavVideoSource("C:\Users\Administrator\Desktop\Nuova cartella (3)\DJI_20250216183449_0077_D.MP4")
SmoothFPS_RIFE(400000, 1000)

Merge(SelectEven(),SelectOdd()) # 200fps 
Merge(SelectEven(),SelectOdd()) # 100 fps
Merge(SelectEven(),SelectOdd()) # 50 fps

#SelectEven() # or Merge(SelectEven(),SelectOdd() again for more blur, 25 fps

ConvertToYUY2(interlaced=false)


AssumeBFF().SeparateFields().SelectEvery(4, 0, 3).Weave()
prefetch(8)

and encode the avs using:

v:\ffmpegOCT2024.exe -y -threads auto -i "C:\Users\Administrator\Desktop\Nuova cartella (3)\DJI_20250216183449_0077_D_HD.avs" -c:v libx264 -pix_fmt yuv422p -profile:v high422 -level:v 4.1 -g 33 -bf 1 -crf 22 -flags +ildct+ilme -x264opts tff=1:ref=1 -an v:\out422.mp4

but I don't know if the result is right, however the encoding speed at 2 fps

[marcorocchini]

I need the final .AVS take an output 1080i50 regardless of whatever the framerate of the source file is

I try the routine above and briefly I seem to notice that the encoding is good: furthermore the final file encoded seems to benefit from a necessary anti-aliasing typical in conversions from progressive to interlaced.


Maybe this is because I'm applying it to the final 50P final output the

AssumeBFF().SeparateFields().SelectEvery(4, 0, 3).Weave()

which brings it to 50i

But I don't seem to see the little "film-look" motion blur effect that I would need. So I wonder: is there a way to get it at the end of the script?

p.s.: at the moment I use model=11 in RIFE to avoid the "only rife-v4 model support custom frame rate" when using model=9

[Sharc]

I need the final .AVS take an output 1080i50 regardless of whatever the framerate of the source file is

I try the routine above and briefly I seem to notice that the encoding is good: furthermore the final file encoded seems to benefit from a necessary anti-aliasing typical in conversions from progressive to interlaced.


Maybe this is because I'm applying it to the final 50P final output the

AssumeBFF().SeparateFields().SelectEvery(4, 0, 3).Weave()

which brings it to 50i

But I don't seem to see the little "film-look" motion blur effect that I would need. So I wonder: is there a way to get it at the end of the script?

p.s.: at the moment I use model=11 in RIFE to avoid the "only rife-v4 model support custom frame rate" when using model=9

Why interlacing it? Leave it as 25p (or if you prefer slow it down to 24p). There is no interlaced film. Film is always a sequence of progressive pictures.
If your player rejects to play 25fps progressive you can encode it with x264 using the "--fake-interlaced" flag.
http://web.archive.org/web/20211016213149/http://www.x264bluray.com/home/1080i-p

[pandy]

I'm a bit confused as it seems you expect something else than your description but anyway:

vf "zscale=f=spline36:r=full:w=iw/4:h=-2,minterpolate=fps=100:mi_mode=2:mc_mode=1:me_mode=0:me=8:vsbmc=1,tmix=frames=4:weights='3 6 12 24',fps=50,interlace=scan=0:lowpass=2"

horizontal resolution reduced by 4 to speedup process - if you are ok then just remove zscale from filterline, 'minterpolate' produce 100fps video sequence so your goal to produce 50i despite source is meet, temporal blur is small - only 4 frames - feel free to enlarge it, weighting scale should be adjusted adequately - also weights are up to you, i prefer geometrical relation but it can be other.
'minterpolate' can be adjusted and/or replaced by some linear filter to speedup processing.

[marcorocchini]

Why interlacing it? Leave it as 25p

http://web.archive.org/web/20211016213149/http://www.x264bluray.com/home/1080i-p

my station requires delivery 1080i25

[poisondeathray]

Why interlacing it? Leave it as 25p

http://web.archive.org/web/20211016213149/http://www.x264bluray.com/home/1080i-p

my station requires delivery 1080i25

ALL 50Hz European stations use "25p in 25i" for "25p" content . Everything is delivered as 1080i25, but there are 2 categories:

eg. ITV, BBC, Sky TV etc... all UK or 50Hz area produced dramas are "25p in 25i" - they are NOT interlaced content. It's progressive 25p material, encoded interlaced or MBAFF

But things like sports, news (not film material, not drama), documentaries, reality TV shows - use real interlaced content 25i - there are 50 different fields or moments in time captured per second, not 25 like the dramas

"Filmic look" is more than FPS, more than motion blur. You need proper lenses, depth of field, shooting technique, grading, grain

[jagabo]

Maybe this is because I'm applying it to the final 50P final output the

AssumeBFF().SeparateFields().SelectEvery(4, 0, 3).Weave()

which brings it to 50i

But I don't seem to see the little "film-look" motion blur effect that I would need.

Of course not. You still get 50 different images per second at playback. To get the film look you need to reduce the frame rate to 25p.

So I wonder: is there a way to get it at the end of the script?

If you want more motion blur apply the Merge(SelectEven(), SelectOdd()) a fourth time (to get 25p) instead of SelectEven(). Then encode the 25p video as if it's interlaced.

"Filmic look" is more than FPS, more than motion blur. You need proper lenses, depth of field, shooting technique, grading, grain

Yes, exactly.

[marcorocchini]

ALL 50Hz European stations use "25p in 25i" for "25p" content

my cat.

ok, but then considering that my delivery should be 1080i50, during the shooting I can only do a few things:

use a camera that shoot 50i but the interlaced format is increasingly falling into disuse. Or use a camera that shoot 25p or use a camera that shoot 50p

Often the shot is in non-ideal conditions, here I assume only the average-worst conditions, where the footage already lacks sufficient conditions to be considered "good" or satisfies the minimum quality requirements.

So

1) shoot in 1080i50 directly and get for example this raw test footage made now in medium-poor conditions, but the rec format is ready-for-delivery

https://www.swisstransfer.com/d/75f01182-3d87-4b9d-b8e8-548b16914aac

i.e. camera that records directly in interlaced format (not progressive) often no possibility to have a tripod, or have if you a tripod you are not very agile and run&gun, gray or ugly sky, camera without internal stabilizer, lens without stabilization system because all 2/3 B4 lenses cannot have stabilizer, presence in the images of dense texture or grid that easily allows moiré or aliasing problems to be generated on the image in post processing

2) have a camera, not very expensive (with at least XLR audio input) that can record in 25p, and I can shoot 1080P25 as here

https://www.swisstransfer.com/d/545b3051-288b-45ba-9d80-979f0a292e09

3) have the same camera that record in 50P, and I can shoot 1080P50 as here

https://www.swisstransfer.com/d/47a2fed4-615e-400c-9f5a-24f9b856d39c

4) manage footage coming from mini cameras like osmo or mini drone that necessarily shoot in progressive, but keeping in mind that the final delivery will still have to be 1080i50 and that in some way these sources will have to be treated appropriately to reduce aliasing or artefacts. Source file like this osmo 25P:

https://www.swisstransfer.com/d/1ea66cd5-0c23-4389-9087-be692c843e0d

or osmo 50P:

https://www.swisstransfer.com/d/f2561263-48f6-4531-a992-6e3c3d8d6132

So, in conclusion:

camera that record natively interlaced are too old and obsolete, however in the local market they can still be reasonably used.

Camera that shot 25P: final result it seems to me that have too "motion blur" when the target output is the 1080i25, even when the shutter speed is set to double the framerate.

Camera that shot 50P: primarily it loses brightness if compared to the 25P and in all cases, 25P or 50P footage, it seems that when the final result have to be 1080i50 a special post-processing is required, or am I wrong?

I'm a confused cat

the basic problem comes from the fact that the delivery in 1080i25 and the use of cameras that record in progressive, 25p or 50p, perhaps requires a pre-processing process on the progressive clips, or I'm wrong. Or the use of true-interlaced and progressive clips in a timeline leads necessarily to edit everything and export in 50P, and then transcode in 25i. Otherwise if the progressive clips are imported directly into the interlaced project timeline, aliasing and moire problems or unwanted artifacts arise. Furthermore, a personal problem of mine is that I always need, always and without exceptions, to use cameras that have an analog XLR audio input to record the audio from the Sennheiser EK3241 receivers. For example, this step would be difficult on mirrorless or cameras mounted or on gimbals

[poisondeathray]

considering that my delivery should be 1080i50

"25i" and "50i" are descriptions for the same thing, different naming conventions. eg. "25p in 25i" is the same thing as "25p in 50i"

camera that record natively interlaced are too old and obsolete, however in the local market they can still be reasonably used.

If it's appropriate for the program, then use it

eg. If you were shooting a scripted drama, interlace acquisition would be inappopriate

Camera that shot 25P: final result it seems to me that have too "motion blur" when the target output is the 1080i25, even when the shutter speed is set to double the framerate.

The topic of this thread was "add motion blur" and make it more "cinematic."

25p (or 24p) "filmic" cinema requires careful planned scripted, controlled shooting with rigs , lighting setups. This for things like dramas and movies . Not news style content, not run and gun

Camera that shot 50P: primarily it loses brightness if compared to the 25P and in all cases, 25P or 50P footage, it seems that when the final result have to be 1080i50 a special post-processing is required, or am I wrong?

Normally shutter speed is 1/50 for 25p , 1/100 for 50p or 25i ( with 25i interlaced content). Faster shutter speed means less light able to hit sensor

You might need special post processing depending on the problems, or the program delivery

e.g. if it was for a drama, but one of the cameras was shot in 50p with 1/100 (not 25p) , you would likely need to discard 1/2 the frames to 25p and add motion blur to simulate the look of the other 25p cameras

Otherwise if the progressive clips are imported directly into the interlaced project timeline, aliasing and moire problems or unwanted artifacts arise.

Aliasing problems arise regardless for any 50p to 25i conversion (for interlaced 25i content conversion, not "25p in 25i" delivery) , you introduce aliasing by converting progressive to interlaced, dropping half the scan lines - this needs lowpass filtering for broadcast to reduce interline twitter, otherwise it will not pass QC . Many NLE's have "reduce interlace flicker" filter. This was covered in your earlier threads years ago - refer back to those for various options. Recall in avs you applied a vertical blur, in ffmpeg you used tinterlace with flags =vlpf


If it's not for some drama, you shoot 50p , or 25i .

50p acquisition is more forward thinking, because you have more possibilities (can be released on other platforms, web) and it should be higher quality . It can be converted to 25i with lowpass

These days UHD acquisition is becoming more common. Some scenarios require it . "1080" is becoming outdated very fast. Think of other scenarios, not just a local TV station - Web, stock video sites, social media

25i acquisition limits your other possibilities. If you are 100% certain it's only for now, only for this station, nothing for future use, then go ahead

[marcorocchini]

ok, this is my horrible drone footage:

https://www.swisstransfer.com/d/d650117a-cfe6-44c7-8874-325830f4a4f9

it is a 25 fps, and if possible I need a transcoded 1080i25 MXF xdcamhd422 format.

For the moment I try this script:

LoadPlugin("v:\BestSource.dll")


LoadPlugin("V:\AviSynth\RIFE\RIFE.dll")
LoadPlugin("V:\AviSynth\RIFE\avsresize.dll")


function SmoothFPS_RIFE(clip source, int num, int den)
{ 
    source
    z_ConvertFormat(pixel_type="RGBPS", colorspace_op="709:709:709:l=>rgb:709:709:f")
    Rife(gpu_thread=1, model=53, gpu_id=0, fps_num=num, fps_den=den, sc=false, sc_threshold=0.10)
    z_ConvertFormat(pixel_type=source.pixeltype, colorspace_op="rgb:709:709:f=>709:709:709:l")
}


BSVideoSource("C:\Users\Administrator\Desktop\drone\25fpsDrone.mp4")


SmoothFPS_RIFE(400000, 1000)

Merge(SelectEven(),SelectOdd()) # 200fps 
Merge(SelectEven(),SelectOdd()) # 100 fps
Merge(SelectEven(),SelectOdd()) # 50 fps

#SelectEven() # or Merge(SelectEven(),SelectOdd() again for more blur, 25 fps

ConvertToYUY2(interlaced=false)


AssumeTFF().SeparateFields().SelectEvery(4, 0, 3).Weave()
prefetch(8)

and encoding using

ffmpeg.exe -v quiet -stats -y -i "C:\Users\Administrator\Desktop\drone\25fpsDrone.mp4.avs"  -an -pix_fmt yuv422p -f rawvideo - | v:\automazioneclip\core\ffmbcX642017.exe -threads 12 -s 1920x1080 -r 25 -y -f rawvideo -pix_fmt yuv422p -i - -tff -target xdcamhd422 -an -ac 4 v:\out1080i25.mxf

attached various tools:
https://www.swisstransfer.com/d/a3ca6fc0-9a1f-4764-9219-2f0210a444b4

out1080i25.mxf maybe appear with a little bit aliasing artifacts. So my question is: how the script have to be changed to get a better result?

and is there a way to speed up the encoding process? maybe even wanting to get a little cinematic motion blur, even considering the low quality source

[jagabo]

For the moment I try this script:

LoadPlugin("v:\BestSource.dll")


LoadPlugin("V:\AviSynth\RIFE\RIFE.dll")
LoadPlugin("V:\AviSynth\RIFE\avsresize.dll")


function SmoothFPS_RIFE(clip source, int num, int den)
{ 
    source
    z_ConvertFormat(pixel_type="RGBPS", colorspace_op="709:709:709:l=>rgb:709:709:f")
    Rife(gpu_thread=1, model=53, gpu_id=0, fps_num=num, fps_den=den, sc=false, sc_threshold=0.10)
    z_ConvertFormat(pixel_type=source.pixeltype, colorspace_op="rgb:709:709:f=>709:709:709:l")
}


BSVideoSource("C:\Users\Administrator\Desktop\drone\25fpsDrone.mp4")


SmoothFPS_RIFE(400000, 1000)

Merge(SelectEven(),SelectOdd()) # 200fps 
Merge(SelectEven(),SelectOdd()) # 100 fps
Merge(SelectEven(),SelectOdd()) # 50 fps

#SelectEven() # or Merge(SelectEven(),SelectOdd() again for more blur, 25 fps

ConvertToYUY2(interlaced=false)


AssumeTFF().SeparateFields().SelectEvery(4, 0, 3).Weave()
prefetch(8)

Once again, you're producing 25i video where you want to produce 25p encoded as 25i.

and is there a way to speed up the encoding process?

You can use a faster motion interpolating frame rate converter. Expect more artifacts and less effective motion blur.

[marcorocchini]

my target is to get an MXF output final file in xdcamhd422 format 1080i25 (interlaced). If I drag and drop a progressive source an place in timeline with project settings 1080i25, automatically the output seems to have aliasing questions or artifacts. I don't know exactly how others work, if they pre-process their sources but at least I tend to pre-process input files other than 1080i25 format that convert the final files that will go into the timeline into 1080i25. Surely this pre-processing requires its own encoding times but once the material is loaded into the timeline, editing proceeds well until the end.

In theory the pre-processing involves a loss of quality since the initial information of the source files is lost in the transcoding process: I know. But it ensures a final conformity between sources that are not of the same type.

Often I have to edit timelines where the main source is MXF in xdcamhd422 1080i25 format, but then many other clips come from different sources such as mp4 from cell phones, videos from youtube, or mirrorless that can run at completely different frame rates and resolutions. The time available is often always short and a pre-processing transcoding takes time away from editing, so this process might seem inconvenient, but otherwise I hear of my friends who get lost in a thousand hells while editing clips with different formats imported directly into the timeline. Surely those who are good will be able to do everything from within the editing program, but unfortunately I can't. So my question would be the same: so for example to obtain a transcoding of my horrible drone at 25 fps, towards interlaced 1080i25 format what would be the right procedure?

[marcorocchini]

can I do the transcoding of my horrible drone

https://www.swisstransfer.com/d/d650117a-cfe6-44c7-8874-325830f4a4f9

25P to 25i

using SVPflow with avisynth?

I have a i9-11900KF and a NVidia 3080

[jagabo]

Sure.

[marcorocchini]

[marcorocchini]

I'm a bit confused as it seems you expect something else than your description but anyway:

vf "zscale=f=spline36:r=full:w=iw/4:h=-2,minterpolate=fps=100:mi_mode=2:mc_mode=1:me_mode=0:me=8:vsbmc=1,tmix=frames=4:weights='3 6 12 24',fps=50,interlace=scan=0:lowpass=2"

horizontal resolution reduced by 4 to speedup process - if you are ok then just remove zscale from filterline, 'minterpolate' produce 100fps video sequence so your goal to produce 50i despite source is meet, temporal blur is small - only 4 frames - feel free to enlarge it, weighting scale should be adjusted adequately - also weights are up to you, i prefer geometrical relation but it can be other.
'minterpolate' can be adjusted and/or replaced by some linear filter to speedup processing.

minterpolate seems very slow, what can be alternative?