Log in

View Full Version : Can you help me with this problem please?


terminator3
7th June 2025, 01:20
thanks for the info, done.

Selur
7th June 2025, 05:24
QTGMC()
sRestore(frate=23.976)
looks fine to me.

terminator3
7th June 2025, 14:54
QTGMC()
sRestore(frate=23.976)
looks fine to me.

thanks

terminator3
10th August 2025, 20:20
QTGMC()
sRestore(frate=23.976)
looks fine to me.

What deinterlacing method do you recommend here? I tried this one and it seemed to work fine, but if I import it into a video program, it seems like the frames are repeating:
AssumeTFF()
QTGMC(Preset="Slower", EZDenoise=0.0, NoiseProcess=0, Sharpness=1.0, SourceMatch=3, Lossless=2, FPSDivisor=2)

https://we.tl/t-0fI3B3PJeg

poisondeathray
10th August 2025, 21:23
What deinterlacing method do you recommend here? I tried this one and it seemed to work fine, but if I import it into a video program, it seems like the frames are repeating:
AssumeTFF()
QTGMC(Preset="Slower", EZDenoise=0.0, NoiseProcess=0, Sharpness=1.0, SourceMatch=3, Lossless=2, FPSDivisor=2)

https://we.tl/t-0fI3B3PJeg

This is interlaced content, remove the FPSDivisor

If you still have problems, post details on the source filter used, and whether or not that was the original stream (it's interlaced content encoded progressively with chroma artifacts likely because of progressive resampling of chroma, instead of interlace aware resampling)

terminator3
10th August 2025, 21:24
This is interlaced content, remove the FPSDivisor

If you still have problems, post details on the source filter used

If I remove the fps divider it stays at 50 fps,

original video:
Formato : AVC
Formato/Info : Advanced Video Codec
Formato del perfil : High@L3.1
Ajustes del formato : CABAC / 8 Ref Frames
Ajustes del formato, CABAC : Sí
Ajustes del formato, RefFrames : 8 fotogramas
ID códec : V_MPEG4/ISO/AVC
Duración : 39 s 360 ms
Tasa de bits : 15,0 Mb/s
Ancho : 720 píxeles
Alto : 576 píxeles
Relación de aspecto : 4:3
Relación de aspecto original : 4:3
Modo velocidad fotogramas : Constante
Velocidad de fotogramas : 25,000 FPS
Estándar : PAL
Espacio de color : YUV
Submuestreo croma : 4:2:0
Profundidad bits : 8 bits
Tipo barrido : Progresivo

poisondeathray
10th August 2025, 21:36
If I remove the fps divider it stays at 50 fps,


50 fps is what it's supposed to be for that video

25 frames/s interlaced is 50 fields/s . That's 50 moments in time per second captured . If you use fpsdivisor=2 your are discarding 1/2 the data and motion will be less smooth


original video:
Formato : AVC
Formato/Info : Advanced Video Codec
Formato del perfil : High@L3.1
Ajustes del formato : CABAC / 8 Ref Frames
Ajustes del formato, CABAC : Sí
Ajustes del formato, RefFrames : 8 fotogramas
ID códec : V_MPEG4/ISO/AVC
Duración : 39 s 360 ms
Tasa de bits : 15,0 Mb/s
Ancho : 720 píxeles
Alto : 576 píxeles
Relación de aspecto : 4:3
Relación de aspecto original : 4:3
Modo velocidad fotogramas : Constante
Velocidad de fotogramas : 25,000 FPS
Estándar : PAL
Espacio de color : YUV
Submuestreo croma : 4:2:0
Profundidad bits : 8 bits
Tipo barrido : Progresivo

That cannot be the original, original, video .

terminator3
10th August 2025, 21:38
50 fps is what it's supposed to be for that video

25 frames/s interlaced is 50 fields/s . That's 50 moments in time per second captured . If you use fpsdivisor=2 your are discarding 1/2 the data and motion will be less smooth



That cannot be the original, original, video .


My confusion, sorry, the original is this:

ID : 1
Formato : ProRes
Formato de la versión : Version 0
Formato del perfil : 422 HQ
ID códec : apch
Duración : 39 s 377 ms
Duración original : 39 s 440 ms
Tipo de tasa de bits : Variable
Tasa de bits : 60,5 Mb/s
Ancho : 720 píxeles
Ancho abertura limpia : 703 píxeles
Alto : 576 píxeles
Altura abertura limpia : 576 píxeles
Relación de aspecto : 4:3
Relación aspecto abertura limpia : 4:3
Modo velocidad fotogramas : Constante
Velocidad de fotogramas : 25,000 FPS
Estándar : PAL
Espacio de color : YUV
Submuestreo croma : 4:2:2
Tipo barrido : Entrelazado
Tipo barrido original : Progresivo
Tipo barrido, método almacenamiento : Campos intercalados
Orden de barrido : Campo superior primero
Bits/(píxel*fotograma) : 5.832
Tamaño de pista : 284 MiB (98%)
Cantidad de pistas original : 284 MiB (98%)
Título : Core Media Video
Librería de codificación : Apple
Idioma : Inglés
Fecha de codificación : 2025-08-10 17:14:52 UTC
Fecha de etiquetado : 2025-08-10 17:14:55 UTC
Colores primarios : BT.601 PAL
Características transferencia : BT.709
Coeficientes matriz : BT.601

poisondeathray
10th August 2025, 21:44
There is a chance the chroma problem is in that prores version too (you have to look at it), but when you convert 4:2:2 to 4:2:0 you should either deinterlace first, or downsample the chroma from 4:2:2 => 4:2:0 in an interlace aware manner while keeping interlace, or keep 4:2:2 interlace

Notice the prores version has "Tipo barrido original : Progresivo" , which is wrong - that can cause editors problems too, you would have to interpret the file as interlaced TFF if using in an editor

terminator3
10th August 2025, 21:47
There is a chance the chroma problem is in that prores version too (you have to look at it), but when you convert 4:2:2 to 4:2:0 you should either deinterlace first, or downsample the chroma from 4:2:2 => 4:2:0 in an interlace aware manner while keeping interlace, or keep 4:2:2 interlace

Notice the prores version has "Tipo barrido original : Progresivo" , which is wrong - that can cause editors problems too, you would have to interpret the file as interlaced TFF if using in an editor

What do you recommend?

poisondeathray
10th August 2025, 21:58
What do you recommend?

It depends on the background information - what are the goals, what is the project, how it's going to be used, archive formats, viewing formats, device targets (compatibility) etc...

terminator3
10th August 2025, 22:23
It depends on the background information - what are the goals, what is the project, how it's going to be used, archive formats, viewing formats, device targets (compatibility) etc...


The objective is to have it deinterlaced, with the maximum quality preserved and that's it, there is nothing more objective.

poisondeathray
10th August 2025, 23:03
The objective is to have it deinterlaced, with the maximum quality preserved and that's it, there is nothing more objective.

Then I have no other suggestions other than deinterlacing to 50p instead of 25p , and encoder to a lossless format (or near lossless if "near maximum quality" is good enough)

People will have personal preferences and debate over different deinterlacers, settings, etc...

Technically there is quality loss when deinterlacing, unless you preserve the original fields

terminator3
11th August 2025, 15:42
Then I have no other suggestions other than deinterlacing to 50p instead of 25p , and encoder to a lossless format (or near lossless if "near maximum quality" is good enough)

People will have personal preferences and debate over different deinterlacers, settings, etc...

Technically there is quality loss when deinterlacing, unless you preserve the original fields

I was referring to the deinterlacing, do you think it's good or not?

poisondeathray
11th August 2025, 21:59
I was referring to the deinterlacing, do you think it's good or not?

It's ok, but really subjective personal taste

eg. I'd personally do things slightly differently and use different setttings

terminator3
11th August 2025, 22:04
It's ok, but really subjective personal taste

eg. I'd personally do things slightly differently and use different setttings

What would you do and use?

poisondeathray
11th August 2025, 22:18
What would you do and use?
I would just keep the original interlaced version archived separately

For a deinterlaced version for viewing, personally I don't use lossless=1 or 2 for QTGMC, because it definitely "looks" worse.

I'd also turn down the sharpness, because it sharpens the noise, predisposes to creating or intensifying existing halos. I'd use border=true, otherwise the top border will flicker more. I'd generally use lower TR settings (to reduce detail loss, and risk of ghosting) and denoise separately for a deinterlaced "viewing" version

terminator3
11th August 2025, 22:44
I would just keep the original interlaced version archived separately

For a deinterlaced version for viewing, personally I don't use lossless=1 or 2 for QTGMC, because it definitely "looks" worse.

I'd also turn down the sharpness, because it sharpens the noise, predisposes to creating or intensifying existing halos. I'd use border=true, otherwise the top border will flicker more. I'd generally use lower TR settings (to reduce detail loss, and risk of ghosting) and denoise separately for a deinterlaced "viewing" version

Please give an example of the code and I'll look at it.

lollo2
13th August 2025, 08:43
Technically there is quality loss when deinterlacing, unless you preserve the original fields

For a deinterlaced version for viewing, personally I don't use lossless=1 or 2 for QTGMC, because it definitely "looks" worse.

Is not present a contraddiction here?

Or in the first sentence you mean that deinterlacing alters the fields to create the new frames (unless you use options to preserve the original fields) and then the "original quality" is somehow lost.

poisondeathray
13th August 2025, 15:23
Is not present a contraddiction here?

Or in the first sentence you mean that deinterlacing alters the fields to create the new frames (unless you use options to preserve the original fields) and then the "original quality" is somehow lost.

Preserving original fields is mathematically lossless (you just discard the interpolated scan lines, and you're left with the original) but subjectively looks worse in terms of flicker and artifacts. You lose a lot of the "magic" - the reason why tempgaussmc_beta2 and QTGMC were created - to reduce flicker and temporal smoothness.

poisondeathray
13th August 2025, 15:34
Please give an example of the code and I'll look at it.

I already made suggestions above

Just play with the settings until you're happy - it's very subjective and depends on your goals

"maximum quality preserved " goal is difficult to define . How do you define "quality"? It means different things to different people. If you mean mathematically lossless original scan lines preserved, it will look worse for 99% of people . But maybe you're in that 1% ?

"closer to the source" is not necessarily a good thing when source has various problems to begin with. If you had a pristine source, already lowpassed, not prone to aliasing, then maybe "closer to the source settings" might be a good choice

Some people like sharpened noise, others do not. Some people hate losing detail, some people hate noise. It's all tradeoffs. Have a look at the documentation and experiment with the settings