16mm transfer formats

I'm new to the forum so still getting the lay of the land. Bear with me.

30 years ago an art school friend of mine shot a 16mm BW documentary on my band. It was at the heyday of the punk scene in Toronto. It was never finished. A few years ago he dumped a box of film and tape off and said "let's finish this thing". We had the film transfered proffesionally to 8 bit uncompressed 4:2:2 SD. The sound was never synched so all the footage was silent. Because it was shot by film students, all the reels were a mish mash of clips that I had to match with a single mono audio track. It was extremely tedious but after a while I had a decent edit. After showing it to a few pros we realized we had something special. They all said the footage was amazing and should be finished and released BUT..... it should be in HD.

So now we are going to rescan the footage in HD. And yes I will have to reinsert all the clips one by one but I see no other way. I'm OK with it.

Question: what format should we get the transfers done to? I will edit in FCP7 and yes the plan is to eventually make DCP file and enter this in some festivals. Ideas?

16mm film contains too fine detail for 720x480 pixels, so it's worth rescanning. But how will you cope with the wide shape of the HD frame? Will you crop off tops and/or bottoms of the 16mm film to make it 16:9? That will reduce the total image detail by one-fourth, as well the ruin some compositions. Or will you pillarbox the 4:3 shaped 16mm image into the 16:9? Is HD that's one-fourth black what the pros meant by "should be in HD"?

As for codec, uncompressed 8-bit and 10-bit 4:2:2 have some archival advantage over ProRes codecs. They will be easily decipherable hundreds of years from now. But they've no practical or artistic advantage today. 10-bits has real pictorial advantage over 8-bits, so you should either scan in uncompressed 10-bit and work in ProRes, or simply scan in ProRes. (Warning: FCP7 handles uncompressed 10-bit poorly.) The ProRes flavor should be 422 HQ, or, only if effects requiring detail in chroma are contemplated, 4444. You could scan in ProRes 4444 (with no alpha) and safekeep it while editing in ProRes 422 HQ.

Pillarboxed 1920x1080 ProRes 422 HQ is exactly the same bits/second as your 720x480 uncompressed 8-bit.

Dennis Couzin
Berlin, Germany

i'll pretty much second what Denis says:

i would Pillarbox the footage,
and chose a ProRes codec.

i did a restoration job recently, and our colourist wanted Prores4444.
some would say that's overkill, but he said for the fine details of what he was doing 4444 would make for a better job.

of course our footage was COLOUR, and yours is BW, so 4444 probably IS overkill.

FWIW, it wont be to long before BlackMagic release their new film scanner, and prices will come way down.
might be worth looking into that.
it will scan at HD and also a frame size double that of HD!
the benefit of that would be if some distributor insists on a full 16/9 frame, you might get a better blow-up,

BUT 16mm is probably at it's limit at HD anyway.
also, im seeing a lot more pillarboxing on TV nowadays, so i think everyone is more comfortable with it.

so my 2¢
Pillarbox,
ProRes422:
wait for Black Magic film scanner , and think about UltraHD



nick

Whoa, you guys are quick! This is exactly what I was looking for. Thank you Dennis and Nick.

I would definately pillar box as I do not want any cropping whatsoever. I might even utilize some of the full frame for a bit of new interview footage I am considering. And yes ProRes422 was exactly what I was thinking even though there might be some colour footage (the interviews).

I forgot to ask about frame rates. I am assuming the scans should be done at 24 fps. From what I understand I am better off editing at 24 fps.

Sure, 24 fps should stay 24 fps.

Sorry I overlooked that the 16 mm was B&W. Then 4:4:4 serves no purpose. In fact, the subsampled chroma information in 4:2:2 also serves no purpose. The ProRes codecs handle this intelligently. A B&W ProRes 4444 file is exactly the same size as a B&W ProRes 422 HQ file, and the latter is just half the size of a color ProRes 422 HQ file.

I recommended ProRes 422 HQ, not ProRes 422. 422 HQ is a sharper, cleaner (less artifacted) codec than plain 422. Especially for 16 mm B&W, where the grain is a part of the aesthetic, 422 HQ is preferable.

Since the image is B&W, pillarboxed 1920x1080 ProRes 422 HQ will be half the bits/second as your 720x480 uncompressed 8-bit was.

Dennis Couzin
Berlin, Germany

hi, Harri

ideally you should do your new scan at the speed it was shot, so it syncs w the audio

but then you already did a telecine (or scan), and synced it all,
so i guess you should scan at the same frame rate as your original scan.

then you can just drop the shots in.

what was your shoot frame rate?
what was your telecine frame rate?


nick

Nick,
The original transfers were done at 29.97 fps. I am going to guess that this might have been a bit of a mistake.

Not sure about the telecine rate. I will be speaking with the person who did the transfers soon and can find out.
Harri

not necessarily,

29.97 and 23.976 are compatible in some ways
frames (or half frames, really) are added to the film to time it out to 29.97

there'd be some way of doing your new work at 23.976.
but it will most likely mean a lot of eye-matchn, which i think you were planning on doing anyway.

i don't work with these frame rates myself, so i cant give solid advice on how to go about it.

the main approach would be to create your master at 23.976,
then make 29.97 versions from that


nick

Quote
Nick Meyers
29.97 and 23.976 are compatible in some ways
frames (or half frames, really) are added to the film to time it out to 29.97

Nick's reference to half frames (fields) shows that his 29.97 is 59.94i. Please, please, all forum members, don't call material simply 29.97 or 29.97 fps when there can any question whether it is 29.97p or 59.94i. If you must call 59.94i 29.97i, OK, just indicate that it's interlaced.

I agree with Harri P. that transfering 24 fps film material to 29.97 fps, probably interlaced, video was "a bit of a mistake". It's not what the film is, and it just makes headaches downstream. Every effort should be made today to discourage presenting film with the poor "film look" of 3:2 telecined 59.94i. See strand from a year ago. Interlace is almost dead. Let it go.

Likewise be done with that 23.976 jive. Multiplying speeds by 1000/1001 is a relic of NTSC analog television, and will just make trouble in the future. Make the 24 fps film 24p video.

Dennis Couzin
Berlin, Germany

[29.97i, OK, just indicate that it's interlaced. ]

Dennis, I haven't yet wrapped my head around the need to do this; I've always treated 29.97 as interlaced, able able to divide cleanly with 23.976. Yet, in preparing a doc for online, they want fields set to None-- that's progressive. Thus, my head explodes.

Is it time for master list of frame rates regarding interlaced and progressive possibilities of all types, and how they're best iused? I'm ready for one.

Best, as always,
Loren S. Miller
www.facebook.com/neotrondesign

30 fps can be either 60i or 30p. 29.97 fps can be either 59.94i or 29.97p. All four animals lurk in the video jungle. While you weren't noticing it, other video people found 29.97p to be a godsend.

I think it's fair to insist that whoever describes their video as 25 fps, 30 fps, or 29.97 fps must indicate whether it's interlaced or progressive. It's much more than a technical detail, like PsF encoding is; it's what the moving picture is.

I think the full list of p and i rates in common use today comprises these 12:
24p, 23.976p,
25p,
30p, 29.97p,
48p,
50p,
50i,
60p, 59.94p,
60i, 59.94i.

There's no way to say how each is best used because this depends on when and where the editor is, what his client's cameras shoot, and how the video will be distributed.

The 24 vs. 23.976, 30 vs. 29.97, 60 vs 59.94 distinctions are overblown. They are trivial except in some cases when you're editing mixed materials. Once you finish a video in, say, 24p, you can conform it to 23.976p with no loss of synch, no visible change of speed, no audible change of pitch. Just a 0.1% change in duration which competent programmers can deal with. So I'd remove 4 items from the list of 12, as being mere technical details.

Dennis Couzin
Berlin, Germany

Very useful, thank you, Dennis.