[tuco]

Slight threadjack here, but...

How exactly does trandscoding to another frame rate work? With 30 fps and 24 fps, really only 6 of the frames in a second should be the same and the rest would all be slightly different moments. So I would think that it obviouslt can reimagine what those other moments were like, so does it just drop 1 of of every 5 frames to make it 24? And is it much harder to transcode between 25fps and 30fps, or is it the same?

I don't know the technical ins-and-outs of the process. Maybe someone else can. But most video formats don't have a key frame for every frame. That would be uncompressed video if it did. So the frames between key frames are pretty much mathematically defined. And I believe motion JPEG just applies its compression to the individual frames themselves instead of having key frames with interpolated frames in between.

So when you convert frame rates into say an editing format, it just reconstructs the video interpolating frames as needed. I believe the only way you'll ever show someone your motion JPEG is on your computer or device in front of you. If you upload it to a hosting site, it will be pretty much transcoded to compressed format that I described.

[Coby]

Small conversions like 24 to 25, or visa versa, are generally done by retiming the video and the audio by a few percent, and the loss in quality is negligible.

24 (or 25) to 30 is as Tuco describes, with the added complication that when 24P is shown on a NTSC display there is "pulldown" added to match the 30frames/60 fields transmission signal. On a DVD sometimes these extra frames are 'baked in' to the mpeg2 stream, sometimes there's a pulldown "flag" that tells the playback device to add them on the fly. So 30P converted to 24P would actually get displayed on an NTSC as 30P again, but now with pulldown .... which will help it match native 24P.

[deltoidjohn]

There's 3 ways to convert 30p to 24p. One method is to conform the 30p footage to 24p. This basically tells the player to play the entire 30p clip at 24 fps. All the frames are kept, but played at 24fps instead of 30fps. Because of the slower framerate, the duration of your clip will increase by 25%.

You described frame dropping, which is a second method of converting to 24p. In this method, various frames are removed from the clip. This method is not very CPU intensive. Your clip duration stays the same, but the motion may end up somewhat jerky depending on which frames get removed. The best case for going from 30 to 24p is to probably drop every 5th frame.

A third method of converting 30p to 24p is to do frame blending. In this method, various frames are blended into a single frame. Like in frame dropping, the duration of your clips will remain the same. Unlike frame dropping, frame blending will give much better transitions than frame dropping when the scene has lots of motion. However, this method is much more CPU intensive.


- Jason

There's a fourth method too: Interpolation. This method is used by some software such as twixtor and AE when using time remapping. Rather than the rather crude method of frame blending, this actually analyses the pixels to predict movement, thereby gererating frames which were not there to begin with. This method works for both increasing or decreasing frame rate (to a certain point).

[deltoidjohn]

I don't know the technical ins-and-outs of the process. Maybe someone else can. But most video formats don't have a key frame for every frame. That would be uncompressed video if it did. So the frames between key frames are pretty much mathematically defined. And I believe motion JPEG just applies its compression to the individual frames themselves instead of having key frames with interpolated frames in between.

So when you convert frame rates into say an editing format, it just reconstructs the video interpolating frames as needed. I believe the only way you'll ever show someone your motion JPEG is on your computer or device in front of you. If you upload it to a hosting site, it will be pretty much transcoded to compressed format that I described.

It's true that some video formats contain only a few full key-frames, but that doesn't mean the other data is just thrown away. The other frames are not just discarded then reconstructed by interpolating the frames as needed, as you put it, rather they are compressed by referencing back to the key frame. All the motion and colour information is still there, it's just included in shorthand. So messing with the time/frame rate/ playback speed during transcoding can often have major effects and usually looks terrible.

It is very difficult to do a good looking 30p to 24p conversion; if it wasn't then we wouldn't have all this hoopla about particular cameras not having 24p. Imagine if you listnened to your favorite song and every 6th beat was skipped, or if they played some beats blended together. Thats essentially what you're doing with a 30p to 24p conversion.

And video codec's with a full keyframe for every frame are not called uncompressed. They are called intraframe codec's and include DV, DVCproHD, AVC-intra, MJPEG, Cineform, and plenty more. Uncompressed video is like taking 24 (or 25 or 30 or 50 or 60) RAW photos per second. It requires massive processing power and storage space - and is unnecassary for all but a few very specialised purposes.

[tuco]

A common approach employed by many when you are doing a lot of manipulation of your video with special effects in the editor is to export it out as uncompressed video and then transcode that into the final format to avoid artifacts.

[jogiba]

The K-5 has 1080p @25fps and no full manual video control . I will wait for a FF Pentax with 1080p60..

[dande]

Yup, the current firmware doesn't support full manual control (to be precise, the movie recording options are the _same_ as in the K-7 with the addition of 1080p). There is a chance that it will be released as a firmware update, but chances are the same for the K-7 too. So I'm sticking with the K-7, and probably switching to 60D or the D7000 in the future.

[Coby]

Well in case any folk haven't been reading up on the K-5 on other forums, the rumoured video specs I listed at the start of this thread are not entirely accurate. For starters, it seems there's no 1080p30, just 1080p25. More importantly, for me at least, there has been no mention of manual shutter control ...

Up until now, I've believed Pentax could be on the cusp of producing a nice Video DSLR, but now I'm convinced their designers just don't "get it" at all - what a wasted opportunity. Bah!

[dande]

Exactly. The hottest trend is video mode, while almost all competitors now have at least manual controls, but some have manual audio controls and many other additions which are mostly software design solutions, Pentax failed to deliver anything to the cinematographer world. In my viewpoint K-5 is nothing more than a K-7 with a better sensor and some improvment in AF speed. So sad, but still hoping, maybe Pentax will announce manual mode firmware for K-7 and K-5 (haha, yeah right)

[deltoidjohn]

A common approach employed by many when you are doing a lot of manipulation of your video with special effects in the editor is to export it out as uncompressed video and then transcode that into the final format to avoid artifacts.

It is common to use intermediate editing codec's such as Cineform and ProRes, espescially for heavy colour grading or compositing. However, these are not the same thing as uncompressed. Uncompressed is used very rarely (even less now that HD has quadripled file sizes) and only for very specific purposes. 1080p @ 24fps = 95MB/second for 8 bit video. HDMI can stream up to 48bit video so uncompressed could be upwards of 1/2 GB/s. Thats about 2 TB for an hour of footage. And you'll need some super fast drives to even play that footage back.

[dande]

And you'll need some super fast drives to even play that footage back.

and even faster computer to work with those files.

[jogiba]

I am glad I have my Panasonic HDC-TM700 3MOS camcorder that shoots 1080p @60 fps to back up my K-7 that shoots 720p30 and 1536x1024 @30fps so I will continue to wait for a FF DSLR from Pentax with at least 1080p30 and full manual control in video mode. I think 25fps here in the US is a total waste IMHO were most are using 24 , 30 and 60 fps.

[tuco]

It is common to use intermediate editing codec's such as Cineform and ProRes, espescially for heavy colour grading or compositing. However, these are not the same thing as uncompressed. Uncompressed is used very rarely (even less now that HD has quadripled file sizes) and only for very specific purposes. 1080p @ 24fps = 95MB/second for 8 bit video. HDMI can stream up to 48bit video so uncompressed could be upwards of 1/2 GB/s. Thats about 2 TB for an hour of footage. And you'll need some super fast drives to even play that footage back.

Yeah, I use ProRes all the time. I was still getting some artifacts doing some special effects. The uncompressed export solved that. No, I don't do hour videos. Yes, the exported uncompressed are HUGE but once done with it, I of course delete it.