converting 25p to 50p

hi there, in order to bring all clips in a project to the same framerate to go to color i have some problems.

by having converted via compressor 25p footage to 50p i reconnected it to the project. Since i can not just simply change the framerate in the timeline to 50p, i try to bring it from a 25p timeline to a 50p timeline by copy and paste the clips. in the 50p timeline the now already 50p clips are loosing about the half of the material. its not speeding up, just croping... how to do this...??? help is needed. thanx pierre

The big question is why 50p if your source is 25?



www.strypesinpost.com

its a mix of formats, most of it in 720p50. and also because of swiss tv, they send docs in 720p50...

Ideally you would convert everything before you ingest for offline editing (you can do a 2:2 pulldown to 50p pretty fast [is it actually a 2:2 pulldown?]). FCP won't let you reconnect without telling you that you have a property conflict.

I'm not sure when FCP looks at timecode or frame number for reconnection, but I believe it takes in point from timecode and the rest via frame number, but I could be wrong.

I can't think of a solution besides to manually extend each shot. You could try to media manage the clips to 50fps and see if MM lets you recompress to a different frame rate, then reconnect with the converted footage. But do a short test before letting it rip on the full doc



www.strypesinpost.com

Make one simple setting in Compressor.
Video: Settings > Motion > Frame Rate: 50.
This makes 50p from the 25p by doubling every frame. Audio is Pass-through and the sound is correct.

Dennis Couzin
Berlin, Germany

I would use Twixtor to create real 50p files - not just doubling the frames. In FCP you can drag a 25p clip to a 50p sequence. The clip will keep it's timing and will have a frame blending, which I personally don't like.
When you create a new 'Twixtored' movie clip you will loose it's TC and reel information which are essential if you want to reconnect those clips in FCP. You can still add those values in FCP (within another project) before doing a re-connect.

Andreas

Apparently what seems to be happening is that he was working with mixed formats, so he converted the 25p footage to 50. Then he reconnected, and FCP trimmed the duration of the clips in the sequence due to a property conflict.

Thing is I don't know how to automate the process from here. Naturally if you can get FCP to ignore the timecode and use frame number for reconnection, the job would be half done. You'll then have to double the frame number for the in/out frame of each clip and adjust the timebase of the sequence. But this is in theory. You may have to look into XML workarounds, and adjust the properties from there.

If your source clips are all 25p (which I don't think is the case), you can export a QuickTime movie from a 25p sequence and convert that with compressor or twixtor.

My suggestion to duplicate frames conversion to 50 is to facilitate the preparation before you edit, or if you have a lot of footage and you don't want to run all that through optical flow.



www.strypesinpost.com

Andreas: when Twixtor converts 25p to 50p does it keep the original 25 frames intact, adding 25 interpolating frames? When Motion uses "optical flow" it creates 50 new frames from the 25, with sometimes unexpected content. (The distinction between interpolation and curve fitting.)

Even with interpolation a problem arises at a straight cut. A new frame interpolating the two surrounding frames may visibly change the cut. So the most intelligent 25p to 50p conversion will recognize straight cuts and either do a simple doubling of one frame or else an extrapolation from one side.

Dennis Couzin
Berlin, Germany

dcouzin Wrote:
-------------------------------------------------------
> Andreas: when Twixtor converts 25p to 50p does it
> keep the original 25 frames intact, adding 25
> interpolating frames?
Yes. Twixtor does in this case. Whenever it meets a frame which matches an original frame the original one is taken.

> Even with interpolation a problem arises at a
> straight cut. A new frame interpolating the two
> surrounding frames may visibly change the cut. So
> the most intelligent 25p to 50p conversion will
> recognize straight cuts and either do a simple
> doubling of one frame or else an extrapolation
> from one side.
You never should use cuts in a slow down interpolation, it doesn't make sense.
Artifacts also may occur on the edges of the video, since the motion vectors suddenly may be wrong. Some patterns or repeating patterns also may cause strange pictures, but this normally won't happen with this fairly simple 25 to 50 fps conversion.
I always use Twixtor with AE as with AE there are more options to avoid those things. Also nesting works way better in AE in my opinion.

From my experience with mixed 25/50 timelines it's always the easiest way to work with 50 fps timelines for a rough cut. Then convert the 25 fps clips used to 50 fps (the complete clip). After it's done reconnect the clips. FCP will complain about TC and reel, but it will correctly map ins/outs to the newly created 50 fps clips.

What Gerard suggested with the XML won't work as FCP for TC information always will take the original data from the clip if available and replace whatever data in the XML with those values. With FCP 7 though there is a little but pretty cool change - if there are no original TC informations it will take the value from the XML.

Andreas

Andreas Kiel Wrote:
-------------------------------------------------------
> Yes. Twixtor does in this case. Whenever it meets
> a frame which matches an original frame the
> original one is taken.

Thank you for this. It made me reexamine Motion's conversion which to my surprise also preserved the original frames. The motion artifacts from Motion looked so extreme that I thought it couldn't be single "off" frames causing them.

Andreas Kiel Wrote:
-------------------------------------------------------
> You never should use cuts in a slow down
> interpolation, it doesn't make sense.

It makes sense when, for example, you try to make a PAL->NTSC conversion by increasing 25 fps to 30 fps. You don't want to take the whole film apart for a conversion.

Andreas Kiel Wrote:
-------------------------------------------------------
> Some patterns or repeating patterns also
> may cause strange pictures, but this normally
> won't happen with this fairly simple 25 to 50 fps
> conversion.

Again, my experience with 25p -> 50p by optical flow is only with Motion, which might be more primitive than Twixtor, but I find almost every shot ruined by large or small telltale morphy details. Clicking through the frames shows a clear sequence: real; fake; real; fake;... Static background is modified by moving foreground objects. Viewers of the displayed video notice this as an amusing shimmer. Not every part of every video should be amusing.

A "repeating pattern" can be a moving hand (with four parallel fingers), etc. The shadow of a ceiling fan drove Motion to madness. I've uploaded this example. I'll try Twixtor to see if it does it better.

By accepting simple frame doubling when converting 25p footage to intercut with 50p footage, you are assured that the result will look no worse than cinema relative to real "real 50p".

Dennis Couzin
Berlin, Germany

dcouzin Wrote:
-------------------------------------------------------
>
> It makes sense when, for example, you try to make
> a PAL->NTSC conversion by increasing 25 fps to 30
> fps. You don't want to take the whole film apart
> for a conversion.
With a conversion from NTSC->PAL or vice versa I would never use any software if it's for professional use. Snell & Wilcox is always the better choice.
But yes, depending on your budget you will get quite good results with Graeme's converter or Compressor - just depends on 'your machines power' and the time you have to do the job.

>
> Again, my experience with 25p -> 50p by optical
> flow is only with Motion, which might be more
> primitive than Twixtor, but I find almost every
> shot ruined by large or small telltale morphy
> details. Clicking through the frames shows a
> clear sequence: real; fake; real; fake;...
> Static background is modified by moving foreground
> objects. Viewers of the displayed video notice
> this as an amusing shimmer. Not every part of
> every video should be amusing.
I've uploaded a real simple 25 to 50 Twixtor based conversion which is more or less basic settings here -- you might compare it to your result with Motion.
I think you'll agree that this result looks a bit better than the optical flow method used in Motion. This doesn't mean that you can't do the same thing in Motion, but you both have to know a lot about Motion's way of work (which I don't have) and have to create multiple masks and layers to make it work.
Twixtor with this simple task is more or less idiot proofed (doesn't mean you're an idiot).
As your example does show quite a lot of compression artifacts it's a good example on how motion vectors can be be handled with a good software. Again -- this doesn't mean it will work any time with any movie. Takes a bit of time to understand what needs to done prior to the conversion.

>
> A "repeating pattern" can be a moving hand (with
> four parallel fingers), etc. The shadow of a
> ceiling fan drove Motion to madness. I've
> uploaded this example. I'll try Twixtor to see if
> it does it better.
see above
>
> By accepting simple frame doubling when converting
> 25p footage to intercut with 50p footage, you are
> assured that the result will look no worse than
> cinema relative to real "real 50p".
I do have a different opinion -- but that's my problem

At the moment I don't have the time to reply in a way which explains 'repeating patterns' or 'sliding objects' in way which helps. But I hope to find the the time in two or three weeks or maybe in 1 or 2 month. Send me a reminder.

Regards
Andreas

P.S.
I don't have any business relation to RE:Vision, I don't have any review or free copies of Twixtor. I'm just a normal customer with maybe a bit more support/bug requests (sometimes) than other customers do have. I've always been happy with the support, bug fixing and changes.

Andreas Kiel Wrote:
-------------------------------------------------------
> I've uploaded a real simple 25 to 50 Twixtor
> based conversion which is more or less basic
> settings here -- you might compare it to your
> result with Motion.
> I think you'll agree that this result looks a bit
> better than the optical flow method used in
> Motion. ...

Wow! Twixtor's basic result is vastly better than Motion's.
Thanks for showing this.

Dennis Couzin
Berlin, Germany

Hi all,

I just uploaded a side by side example of what dcouzin * uploaded and I did using Twixtor/AE.
Left original 'frame doubling', mid Motion, right Twixtor/AE. You should go thru the movie frame by frame to see the differences.

As said I used the sample without any special tweaking except the speed. Obviously the gamma differs between AE and FCP, the astonishing thing is that the gamma in the FCP->Motion->FCP differs as well.

[www.spherico.com]

Regards
Andreas


*) btw I hate hate these names like 'dcouzin' or anything like that - why not a real name

Quote
Andreas
I just uploaded a side by side example of what dcouzin * uploaded and I did using Twixtor/AE.
Left original 'frame doubling', mid Motion, right Twixtor/AE. You should go thru the movie frame by frame to see the differences.

Hmm, interesting...

I just took the original file that Dennis posted (beadchain 25p.mov) and ran it through Compressor with Frame Controls turned on, proper timing settings, and 50 fps specified in video codec setting window.

My result looks really close to Andreas' Twixtor output and way different that the "mid" Motion output that is in the 3-up comparison movie that Andreas posted. So, I wonder about the Compressor settings that were used...


-Dave

I would like to point out that between Motion, Compressor and Shake, they use different implementations of optical flow and the results vary. Between Compressor and Motion, there has been instances where the footage completely falls apart (garbled pixels) while Compressor was still able to run a decent conversion, albeit with a few artifacts. I have known of Shake guys who prefer using smoothcam and time warp in Shake as opposed to Motion because they get better results.



www.strypesinpost.com

Replying late to D-Mac and strypes.
Strypes is right that Compressor's optical flow is different from Motion's, and D-Mac is right that Compressor does a respectable job on doubling the number of frames in beadchain 25p.mov when "rate conversion" is set to "better" or "best".

beadchain 25p.mov is actually a poor sample for testing rate conversions because its motions are very simple. Motion creates its own problem in this sample when it confuses a flickery shadow filling the background with the beadchain in the foreground.

I tried Compressor vs. Motion on optical flow slowing of more complex material involving crossing objects. Their results looked awful, each in its own way. Compressor's looked strobic. Motion's looked slimy and "overcreative". Surprisingly Compressor Better and Compressor Best produced almost identical results and took about the same time to run.

I'll try Twixtor on the more complex sample and report.

Dennis Couzin
Berlin, Germany

Quote
Andreas Kiel
Obviously the gamma differs between AE and FCP, the astonishing thing is that the gamma in the FCP->Motion->FCP differs as well.

Starting with a greyscale bitmap, I used FCP to create an 8-bit unc 4:2:2 clip, and exported this as a .mov file. I imported this file into Motion and then exported from Motion a new 8-bit unc 4:2:2 .mov file. I also brought the first file into Compressor and created a new 8-bit unc 4:2:2 .mov file. Playing these three files with QuickTime they looked identical. Examining the files with a binary editor the Y' values in the three are like so:

FCP / Motion / Compressor
16 / 16 / 16
33 / 33 / 33
57 / 57 / 57
84 / 84 / 84
118 / 118 / 118
149 / 149 / 149
179 / 179 / 179
208 / 208 / 208
231 / 231 / 231
247 / 235 / 247
254 / 235 / 254

There is no gamma difference. The difference is that Motion clips the super-white values to 235 while Compressor preserves them. This difference was not visible via the QuickTime player.

But I think I have noticed visible differences between the FCP, Motion, and Compressor clips (when following the same routine described) except with ProRes as the codec. That would be astonishing.

Dennis Couzin
Berlin, Germany

Gamma... it doesn't really differ in AE. At least it's not supposed to. When it does, it usually means that it is getting a QT decode to RGB. That QT decode is noisy as hell.

It's not surprising that Motion clips highlights. Compressor used to clip early for some conversions too, but I think it's fixed now.



www.strypesinpost.com