Super 8 film, eliminating "pull down" effect

[2Bdecided]

...but I can watch the NTSC DVD of Finding Nemo (for example) on an NTSC TV, and at the same time watch the PAL DVD of Finding Nemo on a PAL TV (though they won't stay in synch for long! Need to keep pausing the PAL one to allow the NTSC one to catch up) - and on panning shots there's no extra blur on the NTSC TV - it just judders irregularly (3-2)! As opposed to the PAL TV, which judders regularly (25p-in-50i = 2-2). (I'm talking about CRTs here - I don't trust flat screen displays for judging motion portrayal).

But as you say, we're way off topic, and not helping anyone.

I wonder, if you were watching my demo, if your eyes would see what my eyes see. They probably wouldn't be able to get over the 50Hz flicker! It's quite rare to find a 50Hz capable CRT in the USA, whereas 60Hz capable CRTs are quite common over here, and 60Hz capable flat panel TVs are ubiquitous.

EDIT: I have learnt something though - I didn't know US networks re-timed films to fit more adverts in! You're probably right than some of the things I've seen are down to this. But that's not happening on the NTSC DVDs and test footage I have.

Cheers,
David.

[jagabo]

...but I can watch the NTSC DVD of Finding Nemo (for example) on an NTSC TV, and at the same time watch the PAL DVD of Finding Nemo on a PAL TV (though they won't stay in synch for long! Need to keep pausing the PAL one to allow the NTSC one to catch up) - and on panning shots there's no extra blur on the NTSC TV - it just judders irregularly (3-2)!

I agree. NTSC 3:2 pulldown (or frame repeats for progressive displays) does not blur the picture, it adds a judder on top of the inherent jerkiness of 24 fps film. If your computer monitor is running at 60 Hz watch the sample video (24v30v60.avi) in this post:

https://forum.videohelp.com/threads/307004-Best-framerate-conversion-%28eg-23-97-to-30-...=1#post1888926

[filmboss80]

Finding Nemo remained in the digital realm all the way to DVD creation. There was no celluloid acetate motion picture film involved in any stage of that transfer process. (Film prints were struck for theater exhibition only.)

Again, we are talking about 2 different things. And again, no one is disputing that there is very visible judder in digital 24p playback on many 60i / 60Hz display devices. But that is not what I was talking about, in addressing the concerns of the OP.

For the most part--especially when mixing film and video sources--other types of productions utilize the traditional telecine process, involving the projecting of 24fps cine film through a 5-blade shutter into a 29.97fps video imaging device. That process creates an OPTICAL 3:2 pulldown effect, blurring some frames together. Inherently at that level, it cannot and does not create judder.

Unfortunately, the term "telecine" has evolved and is used to include 24p digital video...and it is in the playback of 24p video that judder occurs.

Jagabo linked to a thread about 24p digital video. The .avi example of the horizontally moving circles was a digital creation. It was not shot on movie film. It is NOT a film-to-video transfer.

In the case of Finding Nemo and many, many other digitally-produced movies today (Toy Story trilogy, Star Wars trilogy included), they go straight from digital master to DVD without going to film. I repeat, there is no cine film to video process involved. In fact, they likely convert the same digital source to both 24p and 25p, thereby introducing time discrepancies between NTSC and PAL.

Again, there is a distinction between 24p digital video (which may or may not have film as its source) played back on 60i/60Hz displays--WHICH DOES CREATE JUDDER--and the optical process of transferring 24fps motion picture film to 29.97fps video, which DOES NOT CREATE JUDDER...but may have motion smearing.

EDIT: If you don't understand what motion picture film is, read this: http://en.wikipedia.org/wiki/Motion_picture_film

[jagabo]

For the most part--especially when mixing film and video sources--other types of productions utilize the traditional telecine process, involving the projecting of 24fps cine film through a 5-blade shutter into a 29.97fps video imaging device. That process creates an OPTICAL 3:2 pulldown effect, blurring some frames together. Inherently at that level, it cannot and does not create judder.

I've never seen that on a DVD. The closest thing I've seen is PAL to NTSC transfers with field blending. Judder is slightly alleviated but you get an annoying strobing as the video alternates between clear and blurry (blended fields) several times a second.

[filmboss80]

24p digital video played on NTSC TV = occasional judder and motion artifacts;

24fps motion picture film to 24p DVD playback on NTSC TV = occasional judder and motion artifacts;

24fps motion picture film, mechanically telecined to 29.97fps video (via 5-bladed shutter) on NTSC TV = smooth motion with occasional smearing;

Anything PAL to NTSC or NTSC to PAL (regardless of source) = occasional judder, flicker, and motion artifacts.

Note that DVD is not motion picture film. Note that 24p digital video, shot with a camcorder, is not mechanically converted to 29.97fps video via a motion picture film projector with 5-bladed shutter.

I am sorry that I have been such an epic failure at helping some of you understand the differences in this thread thus far.

Try to think back to the analog age, and how Hollywood movies looked on your VHS tapes. Was there judder? Stop-start sticky frames? Mechanical telecine back then did not cause such things. The causes of judder and motion artifact in DVD, Blu-ray, and digital television are attributed to many things -- but not motion picture film transfer through a 5-bladed shutter.

This forum thread was simply about the process of getting a movie film source--NOT videotape, DVD, or digital video file sources--onto a digital format, while eliminating the flicker caused by consumer movie projectors (which do not have the 5-bladed telecine shutters).

Sheesh!

[jagabo]

Try to think back to the analog age, and how Hollywood movies looked on your VHS tapes. Was there judder?

Yes. The majority of film sourced VHS tapes I've dealt with have been normal 3:2 pulldown. Like 95 percent of film based TV broadcasts -- both before and after going digital.

Pleas supply a sample of a 5 blade shutter transfer.

[filmboss80]

Try to think back to the analog age, and how Hollywood movies looked on your VHS tapes. Was there judder?

Yes. The majority of film sourced VHS tapes I've dealt with have been normal 3:2 pulldown. Like 95 percent of film based TV broadcasts -- both before and after going digital.

Pleas supply a sample of a 5 blade shutter transfer.

All your theatrical movies on VHS provide endless samples, as well as many film-based television broadcasts (though some of the latter are subject to digital time compression and will have judder). I NEVER stated that they did not have 3:2 pulldown. All do. I have stated that from the beginning. (Does anyone actually take the time to read what has been written?) My explanation was that the optically-generated 3:2 telecine effect (not to be confused a digital player's handling of a 24p video) is one which produces motion smearing but not jerky, stop-start stutter effects.

There is no way to get perfect film transfers in any setting, but there are ways that you can achieve a film-to-video transfer, free of flicker and jerkiness, albeit at the cost of motion smearing (which is the outcome of optical 3:2 pulldown).

I don't know how many times I need to repeat it. The weekend beckons; so I'll wait a few days to see what other misunderstandings and misinterpretations will need to be addressed.

Adding on: Ah, here's a link that depicts differences in 5-bladed shutter transfers and 2k scans of films: http://framediscreet.com/transfers/
The bottom sample is a good indication of how there is blurring in the motion. The right side scans (not from 5-bladed shutter transfers, but scans of individual frames for 24p playback) are much sharper and cleaner, but that process is where you're getting judder. There is always a trade-off. (Aerial imaging, in which the camera's CCD directly shoots the film gate, will create much sharper 5-bladed shutter transfers than the examples given.)

Unfortunately, these are still frames, but most of you can do your own motion comparisons with VHS movies and DVD movies.

[jagabo]

Try to think back to the analog age, and how Hollywood movies looked on your VHS tapes. Was there judder?

Yes. The majority of film sourced VHS tapes I've dealt with have been normal 3:2 pulldown. Like 95 percent of film based TV broadcasts -- both before and after going digital.

Pleas supply a sample of a 5 blade shutter transfer.

All your theatrical movies on VHS provide endless samples, as well as many film-based television broadcasts

No they don't. I've rarely seen anything other than straight 3:2 pulldown. No blended fields or frames. I have looked very closely at hundreds. (Some have missing frames or fields from time compression, of course, but that is a separate issue.) About the only time I've seen blended fields/frames are with PAL to NTSC conversions.

[filmboss80]

It is the weekend; I don't have much time, but I'm determined to try and make you understand about film-to-video transfers utilizing telecine film system with 5-bladed shutter. Have a look at this nice diagram:



Notice that Video Frame 4 is made from the BLENDING of Film Frames 3 and 4. This is what I've been trying to explain.

Perhaps, I simply have no credibility with you. So, listen to others...

From http://www.digitaltransfersystems.net/technologycomparison.aspx under the section on Aerial Image Transfer (Real-time):

This method became prominent early this decade as DVD was becoming popular and consumers expected better quality for their transfers. This technique utilizes a special projector which has been modified with a 5 blade shutter to eliminate flicker. The film frame is projected onto a mirror which is enlarged to facilitate a video camera (typically 3CCD) to capture the resultant image through the condenser lens.

The results are usually flicker-free but present blended frames which affects the sharpness of the image.

Another telecine primer, http://simplydigitalvideo.com/TransferMethod.htm , offers this:

And lastly a video frame may actually consist of a partial exposure of two different film frames. When a video frame is being "exposed", the projector may pull down the next frame. Thus producing a double exposure.

You DO see frame blending, only it simply appears identical to the double image that occurs when combining video fields of a motion sequence. It is hard for the eyeballs to determine whether each field of a video frame is taken from a single film frame or if each video field came from separate film frames. However, you cannot get 30 frames of video from 24 frames of film without some film frames being blended.

Honestly, I'm doing the best I can to help you understand. This is all that I can do for now.

[jagabo]

It is the weekend; I don't have much time, but I'm determined to try and make you understand about film-to-video transfers utilizing telecine film system with 5-bladed shutter. Have a look at this nice diagram:


[Attachment 5901 - Click to enlarge]

Notice that Video Frame 4 is made from the BLENDING of Film Frames 3 and 4. This is what I've been trying to explain.

That is nothing more than standard hard telecine with 2:3 pulldown. Nothing is blended, the two fields (3.1 and 4.2) are simply woven. Normal telecine like that judders on playback on 60 Hz devices, interlaced and progressive.

[manono]

However, you cannot get 30 frames of video from 24 frames of film without some film frames being blended.

That's simply not true. I've been keeping out of this thinking you were actually going to show us something, probably film field-blended to 29.97fps which only the worst fly-by-night home video production companies use instead of soft or hard telecine. But no, you insist on claiming that regular hard telecine is really something else. And it's not. Nothing is blended. Fields are only repeated/duplicated.

[2Bdecided]

Finding Nemo remained in the digital realm all the way to DVD creation. There was no celluloid acetate motion picture film involved in any stage of that transfer process. (Film prints were struck for theater exhibition only.)

I also have Little Giants, Richie Rich, Wyatt Earp, Free Willy, Outbreak, Batman Forever, and The Fugitive.

All on both PAL and NTSC DigiBeta - all hard telecined. Most on PAL and NTSC commercial DVDs - all different transfers from the DigiBeta. Plenty of other PAL and NTSC DVDs and VHS tapes (though most of the VHS test tapes are worn out now).

Anyway, jagabo and manono clearly get it too, so I'll leave you to it.

Cheers,
David.

[2Bdecided]

FWIW the only telecine machine I've heard of that will produce blended fields is the polygon.

http://lmgtfy.com/?q=polygon+telecine

(7th hit includes interesting information about its use with silent films - though way back when it was also commonly used as a bog standard telecine, free running, as described in the first hit)

Cheers,
David.

[filmboss80]

Returned to this thread amidst a firestorm…inadvertently of my own making.

In the film community (but apparently not so much in the digital video community), “frame blending,” has been (at least in the past) an accepted term associated with the intermittent, 2-combined image phenomenon produced when a realtime telecine projector with 5-blade shutter optically emits 120 flashes of light while projecting 24 frames of film (5 x 24 = 120), as those optical flashes are captured by a device with a shutter speed set for capturing 30fps (more precisely 29.97). Whether you have a 60i or 30p video recording device at the end of the telecine’s aerial imaging system, two out of every five full video frames will be images derived from 2 different film frames. In the case of interlaced video, one film frame can be isolated to the first video field and the successive film frame can be isolated to the second field. With real-time transfers utilizing projectors with 5-bladed shutters, you’ll have some film frame sequences enduring 2/60ths of a second (equaling 2 NTSC video fields or one NTSC video frame) and some enduring 3/60ths of a second (equaling 3 NTSC video fields or 1.5 NTSC video frames). When recording at 30p, how should one describe the image seen on a progressive video frame that captures the last 1/60th of one film frame and the first 1/60th of the succeeding film frame? I was fine with “frame blending,” even when capturing on interlaced video (and I’m far from being alone in using it), but if the term is never okay to use in this forum, I am certainly open to whatever you folks prefer to call it.

Lest I create further misunderstanding, allow me to clarify that I in no way imply that a 5-bladed shuttered telecine projector somehow plasters 2 film frames together or superimposes one over the other. It is purely an optical image phenomenon that occurs when a camera with a 1/30th of a second shutter captures 1/60th sec. of one film frame and 1/60th sec. of the succeeding film frame. All the same, it is never the projector alone that causes the blended film frame effect to occur. In fact, if you remove the video capture device from the equation, you’ll optically have 120 flashes of light per second, and at no point does any single light flash contain 2 film frames. It is only when the NTSC video imaging device enters the equation that the blended film frame effect occurs—and it is seen in full video frames (be they 2-field interlaced or 1-frame progressive) twice in every 5 video frames. That was what I was trying—though failing miserably—to communicate back in Post #28.

I did not intend to imply that a completely different form of telecine pattern was being generated; only that its unique combination of optical film frame display and electronic NTSC video capture created a motion aesthetic unlike frame-to-frame scanning of film to 24p video.

Another thing to clear up: nowhere did I assert that the use of a 5-bladed-shuttered telecine projector offered the best possible film-to-video transfer result. Furthermore, it is neither the preferred of current method of film transfer employed by my place of business. (Certainly, I have done countless transfers that way in years past, but not for quite some time.) Back in Post #5, I wrote that it was merely one out of five possible options available to the OP. I also offered this link: http://framediscreet.com/transfers/ which illustrated the deficiencies of the technology. In a later post, I specified its usefulness in projects that incorporate both film and 29.97 video. (In that application, there’s no practical way of avoiding hard telecine.)

On the issue of judder, when talking about VHS movies in an earlier post, I initially spelled out that: “…there was no perceptible judder or motion blur--although technically, there was plenty of both…” In later posts, I failed to keep acknowledging the definite existence of telecine judder. I also neglected to repeat that my opinions on judder applied specifically to average viewer perceptions. I have no problem issuing a mea culpa for the way I stated things at the end of Post #33. I had intended to assert that the aforementioned film-to-video method in no way produced the kind of judder seen in the 24v30v60.avi clip in https://forum.videohelp.com/threads/307004-Best-framerate-conversion-%28eg-23-97-to-30-...=1#post1888926 . In hindsight, that one admittedly reads like an adherence to some notion that film-to-videos can be made completely judder-free. My bad. I willingly eat crow for the way that one came out.

Wow, I’ve written tons, and yet I fear that I may not have fully addressed every single inferred heresy. I’ll be back in a few days. If I still need to clear the air further, let me know.

Until then, much love.

[jagabo]

I had intended to assert that the aforementioned film-to-video method in no way produced the kind of judder seen in the 24v30v60.avi clip in https://forum.videohelp.com/threads/307004-Best-framerate-conversion-%28eg-23-97-to-30-...=1#post1888926 .

Burn that video as a 30i DVD. If you do it properly you'll see the same 3:2 pulldown pattern on the top row of circles if you look at individual frames. And it will judder just as much on a TV as the original does on the computer. Yes, I used worst case test pattern to make the problem obvious. But once it's pointed out to them, even the average viewer can see it in "normal" video too. Certainly 2bDecided was seeing it.

Software can produce hard telecine 3:2 pulldown from 24p sources exactly like a five blade projector.

[Khaver]

Here's a utility for extracting a film capture using a camcorder and a projector with the shutter removed and running at a slow speed (4-10fps). Camcorder should be 50i/60i with a shutter speed around 1/100 sec.

It's a compiled AutoIT program. It uses avisynth and some plugins, ffdshow and ffmpeg. Docs included.

[szilagyic]

I just wanted to follow up on this post with some of the results of the project. I ended up getting a GAF 3000S projector that has 18 & 24 fps speeds, with a speed adjustment knob. An excellent projector and very versatile. I ended up setting the video camera to 24 fps (it has a special "cinema mode" setting -- Panasonic HDC-TM700). I switched the projector to 24 fps, adjusted the speed to match the video camera and started capturing. I noticed that when adjusting the speed to match the video camera at the beginning of the reel is good until about halfway through the reel and I have to adjust the speed and focus on the projector again (and slow it down very slightly). I did have to leave the "iris" on the video camera to automatic, which allowed it to adjust to the light/dark scenes in the film. It looks similar to recording with a camcoder when you pan from a light to dark area, etc.

After the reel is captured, I am using AVISynth to slow the speed of the video file down to the correct speed. First I am using the "DGIndex" utility to save the D2V file to save the audio to a separate WAV file. Then the WAV file is opened and cleaned to remove the hum from the projector motor. A sample AVISynth script I am using is:

LoadPlugin("c:\program files\avisynth 2.5\plugins\DGDecode.dll")

video=MPEG2Source("T:\movie_project\reel_01\avisyn th.d2v")

audio=WAVSource("T:\movie_project\reel_01\avisynth _proc.wav")

AudioDub(video,audio)

AssumeFPS(21000, 1000, true) # Convert frame rate, also adjust audio.

Note the fps in the AVISynth script is 21. This was what I selected to obtain the most accurate audio since the film was captured at 24 fps "cinema mode" (30 fps NTSC video) but the film was originally recorded around 18 fps. I open the AVS file in VirtualDub, then apply the "deinterlace" (blend), and "sharpen" video filters. I also change the audio sample rate to 48000 Hz, high quality. I export an uncompressed AVI file from VirtualDub, then compress it to Mpeg-2 files (at 24 fps -- the AVI files are 21 fps so I figured 24 fps is for film and would be the closest standard possible) then burn to DVD.

So far I am very pleased with the results. I know there are better ways to capture old film such as a telecine, however with this setup I am getting very good results and it looks pretty good playing back on a regular TV.

I appreciate all of the feedback and guidance, and hope this helps anybody else taking on a similar project!!

--
Chris

[2Bdecided]

If you're really using the 24p mode on the Panasonic HDC-TM700, you shouldn't need the deinterlace step in VirtualDub - the video isn't interlaced, it's progressive.

If you deinterlace (blend) an already progressive source, you simply apply a gentle vertical soften to the image.

Also, if you're really using the 24p mode on the camcorder, then to slow down 24fps to 18fps, you'd ask AVIsynth to output 18fps.

So maybe you're using 30p or 60i instead? Or maybe I don't really know what the Panasonic HDC-TM700 does in 24p mode (though I know what it should do!).

Cheers,
David.

[szilagyic]

If you're really using the 24p mode on the Panasonic HDC-TM700, you shouldn't need the deinterlace step in VirtualDub - the video isn't interlaced, it's progressive.

If you deinterlace (blend) an already progressive source, you simply apply a gentle vertical soften to the image.

Also, if you're really using the 24p mode on the camcorder, then to slow down 24fps to 18fps, you'd ask AVIsynth to output 18fps.

So maybe you're using 30p or 60i instead? Or maybe I don't really know what the Panasonic HDC-TM700 does in 24p mode (though I know what it should do!).

I initially thought the same thing although when using the cinema mode and setting the shutter to 24 fps, but there is no progressive mode that I've been able to find. I am using the HDWriter software that came with the camera that converts the AVCHD videos to "standard videos" which is for sure 30 fps and interlaced. I've verified this because I need to slow the video down with AVISynth to about 21 fps and that gets the sound right on. I initially tried 18 fps and the sound was clearly too slow. I've looked through the manual, and the only progressive mode is the 1080/60p mode that it has which does not allow you to use the 24 fps cinema mode.

I really appreciate the feedback. If I'm missing something with the camcorder settings please let me know. Thanks!!!

[jagabo]

I initially thought the same thing although when using the cinema mode and setting the shutter to 24 fps, but there is no progressive mode that I've been able to find.

24p is often encoded by cameras as 30i after 3:2 pulldown. You have to inverse telecine to get back to 24 frames. Is there any way you could upload a short cinema mode clip with significant motion? Say, a medium speed panning shot (fast enough so the motion is very obvious, not so fast that everything is blurred).

[poisondeathray]

TM700 doesn't shoot native 24p ; it's hard telecined 24p in 60i

[vhelp]

this is all new to me so excuse me if i misunderstood anything in this 8mm-to-video transfer endeavor.

you can not shoot that 8mm with just any video camera, whether 24p, 30i, 60i, etc. without the complication of the "strobbing" the "dark wave" you mentioned in your original post which you already know if due to mis-match frame rates. even using 24p via camera's telecine technique (already mentioned above) is still going to give you the strobbing problem. avisynth is not going to remove it--if at all--because the strobe (dark wave) was captured (recorded) by the camera.

you have to use a video camera that features an 18 fps frame rate, but i don' t know of any that do--its not standard in todays equipment. although you can force a "capture frame rate" to 18 fps (virtualdub for instance) you still need a camera that can "record" the source's 18 fps first. so the source of your problem is obtaining a video record and/or capture to true 18 fps. and the only way i know (just thinking about it) is to rig up a DSLR camera via usab cable and software and capture each frame. and you also have to rig something up for the "signal" part, the part that determines the 8mm projectors new frame so it can take that picture. that would give you maximum quality especially if you set the dslr camera to snap pitures in the raw. there are all sorts of ideas and ways to go about this, you just have to sit down and review how your 8mm projector works and think about how you can control it through start/stop functions, electronically--you already have the dslr camer's usb port and software to control said camera--this is something i would love to do as a fun hobby.

the only right way to transfer these 8mm (18fps) videos is to capture each frame manually, then bring each image into a time line or video editor (or avisynth scripts) to restore original 18 fps frame rate and save out as a new video source file. you can even go the extra mile (after creating new 18 fps avi file) by encoding to an mpeg-2 video and then pad the frame rate with flags (i believe) by using the dgpulldown to 30 fps. i'm not sure it support 18 / 29.970 fps or some other combination that would give you same (or near) fluid play, but you can investigate this if its something you want to consider. maybe someone is familiar that tool and can give you a better (accurate) answer to that idea.

[vhelp]

ok, so i was thinking about this some more, and, you know, a thought just occured to me.. you can try this idea:

what you will need:

* hopefully you have a analog (pci or usb2) capture card!!
* virtudldub capture software

1. connect the capture card to the Panasonic HDC-TM700 output (composite or svideo) ports
2. open virtualdub
3. move mouse down to lower-right of virtualdub window and select the default capture fps, either its 25fps or 30fps
4. after clicking that button, select the 19 or 18 fps selection
5. begin capturing, use a lossless codec, huffy, so you get very little cpu usage and no frame drops.

in this excercise, you will probably see a ton of frame drops in virtualdubs right-side info pane. just ignore it. this is a disconnect with virtualdubs ability to lock to traditionsl frame 29.970 rates and frame drop reporting.

what are we hoping to get out of this:

well, in theory, since the projector is 18 fps, and you video camera is 29.970 fps, i am basing on a hunch that if you capture at 18 fps in virtualdub, that you will get the actual 18 fps images, albiet in fields since capture cards on process by fields. other possible ajustments could be to change the fps in virtualdub, not to 18 fps but to 36 fps, manually in the capture settings, type in 36. the only caviet is that the driver for the capture card does not lock out manual fps adjustments. i have a huntch that if not 18 fps does it then 36 fps, remembering that capture cards only capture (return) fields. good luck.

suggestions for where to get cheap capture cards:

bestbuy - pinnacle usb2, $49-$79, hauppauge hvr-950Q usb2, $69
bjs, officemax, bestbuy - roxio usb2, $49, vhs2pc usb2, $49

[szilagyic]

24p is often encoded by cameras as 30i after 3:2 pulldown. You have to inverse telecine to get back to 24 frames.

This is exactly correct.

I follow you so far; I can see your point by doing the inverse telecine, that it will obtain a better picture in the long run and less motion blur which I did notice in some of the videos I created.

I tried using the "decomb" filter in AVISynth to do the inverse telecine, and it reduced the fps back to 24, then I used "AssumeFPS" to slow it down to 18 fps and the sound is right on so it seems to be at the correct rate as it was originally filmed. The results are actually quite good, and in fact seems smoother than the original way I was doing it. The AVISynth script I'm using is:

LoadPlugin("c:\program files\avisynth 2.5\plugins\DGDecode.dll") 
LoadPlugin("c:\program files\avisynth 2.5\plugins\Decomb.dll")
 
video=MPEG2Source("T:\movie_project\test\avisynth.d2v")  
 
audio=WAVSource("T:\movie_project\test\avisynth_proc.wav") 

AudioDub(video,audio)

#Decomb filter
AssumeTFF()
Telecide(guide=1)
Decimate()
 
AssumeFPS(18000, 1000, true)                 # Convert frame rate, also adjust audio. 
#SSRC(44100)                            # Restore audio sample rate to a standard rate.

Now my question is, having the videos at 18 fps now, I will have to encode them to MPEG2 for DVD. Is it better to use 24 fps or 30 fps when encoding to DVD-compliant MPEG2 files?

As always thank you very much for the feedback, the knowledge in this thread has been awesome.

[manono]

Now my question is, having the videos at 18 fps now, I will have to encode them to MPEG2 for DVD. Is it better to use 24 fps or 30 fps when encoding to DVD-compliant MPEG2 files?

You use the DVD compliant framerate closest to the actual framerate, in this case 23.976fps. But before doing that you have a decision to make. The lowest 'real' framerate to which you can apply DGPulldown is 19.98fps. You have to go from 18fps to 19.98fps one way or another (assuming you want to encode as progressive 19.98fps). One way is to use a ChangeFPS command (ChangeFPS(19.98)). That will add about 2 duplicate frames every second, making for a very slightly jerky viewing experience.

Another way, and the way I do it for silent films, is by using the Motion.dll which, instead of creating duplicate frames, creates 2 blended frames every second, making for smoother playback, but with 2 slight blurs every second. The motion.dll is found here:

http://avisynth.org/warpenterprises/

Also, if you go that route, because you have to encode at 23.976fps but later on you'll be slowing the 'real' framerate back to 19.98fps, you increase the bitrates by a factor of 23.976/19.98=1.2. For example, if a bitrate calculator tells you to use an average bitrate of 4000, you really set it at 4800 in the encoder. Do the same for the min. and max. bitrates.

Or you can dupe and hard telecine frames to achieve 29.97fps, a vastly inferior solution if you ask me, but that's the way most of the DVD production companies have handled 18fps silent films. If you go that route, of course you encode as interlaced 29.97fps.

Also, I'd leave the audio out of the script, reencode it as AC3 audio, and bring it back in during the authoring phase. That allows for a higher max bitrate for the video than if you use WAV audio.

[2Bdecided]

TM700 doesn't shoot native 24p ; it's hard telecined 24p in 60i

Oh good grief!

If I wanted to use 24p (which I don't), I'd find that amazingly annoying. I know people had to put up with this on the HV20, but real 24p was available on the HV40. Going back to this stupid idea, especially on such an advanced camcorder (and with non-tape-based media too!) is just so stupid!!!

Shame on Panasonic!

I wonder what the "PAL" model does?

Cheers,
David.

[szilagyic]

You use the DVD compliant framerate closest to the actual framerate, in this case 23.976fps. But before doing that you have a decision to make. The lowest 'real' framerate to which you can apply DGPulldown is 19.98fps. You have to go from 18fps to 19.98fps one way or another (assuming you want to encode as progressive 19.98fps). One way is to use a ChangeFPS command (ChangeFPS(19.98)). That will add about 2 duplicate frames every second, making for a very slightly jerky viewing experience.

Ok that makes sense, sort of. I am looking at the ConvertFPS filter as well as the Motion filter as you suggested. What about using these filters to convert from 18 to 24 fps, then set the DVD MPEG2 encoder to straight 24 fps, progressive frames? I guess I'm not sure if this would have any benefit over the method you suggested by converting to 19.98 fps and using DGPulldown.

Thanks!

[manono]

Well, don't use ConvertFPS. It'll blend the heck out of it. Done right the Motion.dll will create only the 2 blended frames per second (for 19.98fps). You could create a progressive 23.976 (which is what you probably want, not 24fps) video, but that will entail more blending or duplicating of frames (6 blended or duplicate frames per second). And when applying pulldown to a 19.98fps video stream, there's no additional jerkiness created the way there is with 3:2 pulldown. It's 3:3 pulldown and plays smooth as silk (except for any jerkiness created in the source video.

That's just how I would do (and have done) it. Maybe jagabo, pdr, or others, will show up to give you their own suggestions.

[poisondeathray]

I wonder what the "PAL" model does?

I haven't used the PAL model (in Canada), but I've heard it's 25p native 17Mb/s

The 60p (or 50p PAL) mode isn't really AVCHD ; it's 28Mb/s h.264

[szilagyic]

Also, if you go that route, because you have to encode at 23.976fps but later on you'll be slowing the 'real' framerate back to 19.98fps, you increase the bitrates by a factor of 23.976/19.98=1.2. For example, if a bitrate calculator tells you to use an average bitrate of 4000, you really set it at 4800 in the encoder. Do the same for the min. and max. bitrates.

OK I am not very familiar with DGPulldown however I attempted this. When I ran the video through DGPulldown, the resulting .M2V file ends up being the exact same size as the source video, is this correct? I am also wondering if DGPulldown can do 18 -> 24 because there is an old post here, that describes doing just that:
http://forum.doom9.org/archive/index.php/t-69876.html
In this thread, neuron2 says "The max will be 3/2 * 18 -> 27. So DGPulldown with custom 18 -> 25 will be fine.". So I am wondering if we can do 18 -> 24 in my case here. I did another test and DGPulldown took the values OK.

Anyway, I took the resulting .M2V file from DGPulldown, and encoded it to MPEG2 at 24 fps. The quality is definitely better than what I was doing before when comparing the two side by side. I just want to make sure I'm using DGPulldown properly in this case. Thanks again for your help!!!