pupil

Originally posted by: Doctor M
Sigh, I'm wishing I had an X0 Project frame from ESB to compare to.
I know every scene/frame has different color characteristics, but right now I'm looking for some global reference that's better than my best guess.

I anticipated a release early this week, but I'm still unsure if I want to back the 25% yellow off to 15% or not.
The thing is at 25% the opening crawl still looks drab-er than it probably should.

And now I'm noticing that boosting the Cyan brings out some very nice detail in Hoth's snowscapes previously lost in all white regions. It's very nice looking (especially the skys) and it doesn't turn white hallways blue like it would have if I boosted it in ANH. Again the Vectorscope shows that even boosted, Cyan peaks out around 50-60% so it's not unreasonable. But is it right?

Again, sigh, any advice anyone could give me at this point would be great 'cause I'm just spinning my wheels now. I don't want Smurf Wars or Goldie Wars or anything, but a little more color wouldn't be out of place. I just don't know what's right anymore.

Originally posted by: Citizen
pupil, in terms of picture detail there will be very little difference between the PAL & NTSC versions because both are from letterboxed source and are being upscaled, but playback on a normal tv there is a noticable difference, namely the (god-awful) 3:2 pulldown applies to the NTSC version means smooth motion is ruined.

Originally posted by: Doctor M
Pupil: what is your source for the 72.4 cents number? Calculated or from elsewhere?

Originally posted by: Doctor M
In the past (and my future plans) involved using BeSweet's simple 24 fps -> 23,976 FRC preset. Do you know of a program that actually can resample as you've described or are you referring to the kind of process BeSweet already uses?

Basically you can still do all this with Sound Forge (since you are using this already).

1) In the NORMAL case of turning PAL to NTSC, where the pitch is 4.096% higher than the original film, you use the pitch correction plugin in SoundForge, decrease the pitch by 72.4 cents, but DON'T tick the preserve duration box.

2) In this ABNORMAL case of turning PAL to NTSC, where the VHS audio of ANH was pitch corrected in the original mastering process, you use the pitch correction plugin in SoundForge, decrease the pitch by 72.4 cents, and DO tick the preserve duration box.

All BeSweet is doing is example 1. I would say 99% of the time, you would just need to do example 1 (or use the BeSweet method) but with this bizarre example of the PAL ANH VHS audio, you would use example 2.

If doing example 2 in Sound Forge, you will see the pull down menu with different settings for keeping the time duration when pitch correcting the audio, mess around and see which sounds best (but like I say, for this project Doctor M, don't bother, you're best getting the soundtrack off an NTSC laserdisc and fitting this to Moth3rs video after you have NTSC'd it).

I think we are on the same wavelength now, I was starting to get a bit confused there with what you were asking

Originally posted by: Moth3r
I just remembered, back in the day when I used to do a bit of work as a DJ (this would be mid to late '90s) there was a piece of hardware available that did some kind of realtime time stretching/pitch preservation for you. Basically it meant that DJs could drastically ramp up the tempo of a track without getting the "helium" effect on the vocals, and it also compensated for any sudden pitch change if you touched the record during a mix.

Digging around Google came up with this, but not much else.

So what I'm saying is that the technology for pitch preservation did exist back when these VHS versions were released, so it shouldn't be so much of a surprise that it was used. Just makes you wonder how many other PAL titles have the correct pitch?

Originally posted by: Gillean


That is easily explained, and probably should be done before other people make the same mistake!

Take this example. Our film source is 100,000 seconds long, we make that faster by 4.096% which leaves us with a PAL version at 95,904 seconds. Now if we slow the 95,904 version down by 4.096% we end up with a copy that is 99,832 seconds long, not 100,000 seconds. The reason is that 4.096% of 95,904 is obviously less than 4.096% of 100,000.

So the problem was you were thinking in terms of the original source where SoundForge is correctly thinking in terms of the current source. Easy mistake and something I hadn't thought of until you mentioned it! So whenever you are adjusting PAL material to 24fps then if my maths is right then you need to slow it down by 4.2709376043%. Been way to long since I've done that sort of maths!

Originally posted by: Doctor M
Hmm. First let me nit-pick and say that technically PAL audio isn't time stretched, it is time compressed since the runtime is less than the original. It runs faster, hence the increased frequency. Correcting it requires time expansion. For most PAL material, like it or not, that's really the best alternative out there if you're turning it back to NTSC.

Still time stretching, its just whats its called. Also, this isnt what normally would be done to return PAL audo to NTSC audio speed, as the vast majority of PAL audio is just sped up by 4.096%, not pitch shifted or time stretched in any way. So you just need to slow it down (resample it) by 4.096% and its back to the normal speed and pitch again. Its just in this bizarre case we have here where the original pitch was preserved with pitch shifting/time stretching.



I dont think you get what i mean. time stretching does it as one job, with pitch shifting you need to do it in two stages. To get it from PAL to NTSC speed, first slow down the track (resample it) to get it playing to the correct length (4.096% slower), then pitch shift it up by 4.096%.

From what I'm seeing with soundforge though, pitchshifting doesn't seem to be done on a percentage basis making this much more wobbly a problem to deal with.

Originally posted by: Darth Editous
Here's a comparison of the spectra of those two samples (the one from Moth3r's DVD and the one from the Definitive LD, scaled to match widths):



It looks to me like there's more going on than "simple" pitch shifting* - see how the spacing between sounds changes?

DE (*not an audio engineer)

Originally posted by: Citizen

3. Most importantly, if you merely resample Moth3r's audio from 25fps speed to 23.976fps speed the correct audio pitch will be lowered by 4.096% and everyone will have deeper voices.
This is because the PAL VHS audio source used for his release has been sped up from the original 24fps film but they (the VHS producers) didn't just resample it they also preserved the audio pitch, undoing that and keeping the audio as best as possible may be difficult because you'll have to resample whilst preserving pitch, I couldn't find any timestretching tool I was happy with, the ones I tried (proper packages) weren't competant enough to stretch and keep pitch without ruining certain pitches, it would ruin John Williams' score.

Use NTSC audio source if you can, but fitting it to the video source will be very difficult if the NTSC audio source doesn't have video to go with it as a guide to sync them up.

What has not yet been decided is whether the audio track will be a time expanded version of the original (which was from a VHS source), or a native NTSC track replacement. I currently have access to a nice PCM recording from the Definitive Collection thanks to Arnie.d


I was hoping to host the 2nd version of this clip on my own since some people have problems with RapidShare, but it's currently not possible.
This clip is before I started thinking Picture Framing, and the audio is the original time expanded (with BeSweet) re-encoded to a low bitrate AC3 (for internet transfer ease).

Originally posted by: Citizen
Can't use the audio from the French LDs as it's dubbed in French so will be using the audio from the Definitive collection, starting to think lugging the player+discs over to my friends house to capture at silly high quality is too much hassle so will probably stick to my Canopus ADVC-100 for the audio capture at 48khz, it'll certainly be good enough and there won't be any khz conversion needed at least but will filter out any background inteference noise before upsampling to PAL speed & the 5.1 trick.