Interlacing progressive footage – how to really do it

[Zachary Helton]

Finally got what I wanted by stringing two AJA encoders together, going out analog and re-capturing digital interlaced. Problem is it takes up two work stations and I can’t batch for a long overnight encoding slumber party, so I’d still love to figure out a more streamlined method.
Again, what I have is this:
– Format: Apple ProRes 422
– Bitrate: Max.: Undefined / Average: 111 Mbps / Min.: Undefined
– Frame rate (fps): Max.: — / Average: 59.940 / Min.: —
– Encoding profile: Normal
– Width (Pixel number): 1280*720
– Pixel Aspect Ratio: 1:1
– Display Aspect Ratio: 16:9
– Interlacing: Undefined (this is what VideoSpec reports. It’s progressive)

And what I need to end up with is this:
wrapper: Quicktime
Video: ProRes 422 (HQ)
Resolution: 1920×1080
Fields: 59.94i
timecode: drop frame

[Keith Slavin]

Hi Zachary,

This reply probably a bit late for your current problem (decoding, upscaling, interlacing and encoding), but just for your information, isovideo will soon releasing their file-based, GPU accelerated, motion-compensated standards conversion / transcoding server. Which can handle your current job fairly easily, quickly and with high quality.
If you still need help with this, you may contact me at keith@isovideo.com.

[Keith Slavin]

Hi Zachary,

This reply probably a bit late for your current problem (decoding, upscaling, interlacing and encoding), but just for your information, isovideo will soon releasing their file-based, GPU accelerated, motion-compensated standards conversion / transcoding server. Which can handle your current job fairly easily, quickly and with high quality.
If you still need help with this, you may contact me at keith@isovideo.com.

[Paul Russell]

If the source was originated as progressive at 29.97 then you will end up with a repeated field within each interlaced frame, except that one field will be odd, one even, the picture content shall be the same essentially, you cannot alter the timebase of the capture as it were, just of the display pattern. In essence 29.97p footage will not have the smoothness of 59.94i, even if the re-encoded material is technically complient for interlaced timebase.

Think of it this way, if you had telecined film, the source film is only ever going to have the motion of 24p. Thats why film still looks different from video, even with a 59.94i telecine.

If your source footage was 59.94p (rare) then your conversion process would essentially chuck the even or odd lines away for the appropriate field and each of your two fields would have captured a slightly different moment, and so you would have the smoothness of motion expected from broadcast video, as well as the now correct scan pattern, but the key is in the source material. You cannot put in what wasn’t captured.

Another handy tool for these kind of jobs is the free app JEC-deinterlacer, which is often more transparent in use than compressor etc. I’m not keen on the qt or adobe routes as they sometime try to enforce things like gamma correction or certain colour profiles, they can sork well but you really need to be up early and on the ball.

[Ht Davis]

Unfortunately, doing this all at once is almost impossible on current hardware. You’d have to know some programming.

IF you know ame at all, you can have one tower set to do all the conversion with one hardware, and output to a folder on an external or a shared folder, and then on the other machine, have it “Watch” that folder in AME, and simply output to another separate location for the final step (watching a folder allows you to apply an automatic queuing and rendering with a set preset; so once the AME on the first machine finishes–if you use hardware encoders in ame it still works–with the first step, you can apply the second without even being there; and, yes it still requires two workstations).

From what I see:
You start out in 720p. So work with the frame size first. Upscale the frame size with everything else remaining the same. Now make any other adjustments to quality in your favorite editor, and (to do the same to each one you may have to script it) then have it export. I’m unfamiliar with FCP and compressor, but I know adobe has very little on the ball for this in premiere. You might be able to script it for AE, and have it queue up, then render them all out in ME.
With 50 videos, you’ll want to just try one, then apply a gausian blur or the like to it and check a preview. Just a couple of minutes of it would do. Once you got that to your liking, take down the blur specifics and you should be able to create it as a filter for AME somewhere (eh, still haven’t figured that one out). Do the watch folder thing and let it go.
Once you’ve done the frame size, you can play with the interlace. You need to set this to upper or lower field. I’ve found that lower-field is generally a better viewing on tv’s. But you can do it however you want. It’s a preference thing.

[Ht Davis]

Fogive my last. Here’s a better understanding of what you’ll need to do.
First, double every frame. That’s right, double the number of frames, and for every frame, you can do this using AE to copy a folder of images. Duplicate the folder. Place all the images into one folder, renaming one set as necessary. You should be able to render this to a file. Place in a new sequence set to the higher frame rate as you need, but keep the sequence in interlaced format (60i), both of the videos should be in progressive, on two separate tracks. Right click the one and tell it to display itself as the first field you need, do similar with the other file, but set it to the second field. Now you can output this video. You have twice the fields you need and twice the frames. Every field and frame is played twice. How do we undo this? Please If I have to explain everything to you… So far, we’ve only just begun. Nest this video in a proper sequence at 60i, speed it up 200%. Now it will blend every field to it’s solid when you output, but display the fields properly staggered for playback. You’ll be running them at 120hz, which is 60i’z playback framerate. If you have 60 progressive frames, you can roll them as 2interlaced frames, and they should look decent. It just takes time to run the files out. Don’t count on any digital software to do this all for you. You have to render images first, double each one, then put them into two tracks, and designate each to play a single field, then dump it in another sequence and speed it up to 200%, add the audio at the end. The amount of time it takes should be the same. It will blend frames together if it has to… ;p Blending two of the same frame yields the SAME FRAME! It will take time to render the output as it processes each step, but if you start with a folder of images you can get a lot more done. You could resize the images to your desired viewing size fairly easy with a droplet. You could apply a sharpening action to one copy of the images so when they blend it becomes more natural-looking. It’s all about how you want to make your workflow.