PC Workflow for HD to SD with EX1
 

I've been looking around for a reliable workflow for generating SD content(DVD encodeable) from HD content. This workflow would resize, color correct and convert to SD MPEG2 with the smallest conversion losses, maintaining quality as high as possible and converting from REC709(HD) color to REC601(SD) color.

With thanx to Alan Levi, here's what seems to work quite well:

1-Edit, composite and cut in your favorite NLE(I'm using Edius). Stay in HD, and wrangle the color space to generate RGB16-235 output. Generate a high quality intermediate file when ready. I use Canopus HQ, but, Cineform would work as well. The Canopus HQ and Cineform codec works in REC601.

2-Canopus' Procoder3 allows the application of a REC709 to REC601 color space converter filter. The changes are very small, but anyone interested in being precise can do this step.

3-Bring the intermediate file into Virtualdub. Use the resize filter in VirtualDub to convert to SD.

4-Use TMPGenc to encode the MPEG2 file from the VirtualDub SD file.
Under "Environmental Setting" select "Interpolate YUV data from 4:1:1 to 4:4:4". Select CCIR-601 as the equation for color space.

As a closing note, it's becoming increasingly difficult to go from HD to SD without introducing a color space error. SD MPEG2 will generate REC601 color mapping regardless of what mapping you input. In other words, inputting HD(REC709) color space to a DVD encoder will cause loss of contrast in the final images. None of these applications wrangle the color space automatically, but, all of them have preferences to set up the I/O to maintain color space.

You can do that with freeware. No need for Procoder 3 and Tmpgenc and achieve at least the same quality. Colorspace convertions are not necessary, REC709 is an often used standard on many dvds. At least the dvds look like they should. They didn't with Canopus Procoder, though. A stretched contrast and soft but completely artifact free look was what I've achieved with it -terrible. Tmpgenc and Hcenc don't suffer from this problem.


I'll write a HowTo tomorrow.
You'll need megui and HCenc.


regards Dennis

Thanx, Dennis. Will look forward to seeing your workflow.
I used TMPgenc because I already had it and it comes highly recommended as the best MPEG encoder around. VirtualDub is free, and has (arguably?) the most artifact free resizing algorithms. As for intermediates, Sheervideo, HuffyUV, PicVideo...any of these will work. Each one has its own way of wrangling the colorspace conversion. It's a fact that the chroma differences, for all intents and purposes can be ignored. But the Luma differences can really look bad if not wrangled in the right way.

This guide seems to work well also.

http://www.kenstone.net/fcp_homepage...put_young.html

Only after an illegal install of Mac OSX on a standard PC :D


regards Dennis

Bill,

I am an Edius user as well.

I am surprised that Procoder is bested by TMPGenc.

My experience using only Procoder Express for SD material going to DVD has been stellar.

Has Procoder fallen behind in the HD space?

I have heard this on the Edius forums as well.

Tim...

I'm fairly new to Edius, so I can't really comment on PC's quality. The process I outlined, above, comes from the Edius Forum, itself. After running some files thru, I'm pleased with the outcome.

A related question: I want to be able to mix footage from my HD100 and EX1. I ran a test where I used 720p footage from the HD100 and 1080p footage from the EX1. I edited them together in FCP (in a XDCAM 1080p sequence) and then exported them as a clip through Compressor and burned a DVD. It looks fine. When I put them on a timeline in FCP together using your method, I can't get it to work right. The 720p clip comes through all zoomed in whereas the 1080p clip is fine. I'm obviously doing something wrong and I'm sure I'm not explaining it well. I'm pretty happy with the image I get doing it my way, but if there's a better way of doing it, I'm all ears.

Thanks.

Bill...

There's a lot of places in the chain where this might be happening. I know VirtualDub can be a little tricky to setup the resize filter.I'll attach a screenshot of my settings for you to try. Note that this is setup for a widescreen 16x9.

I never even got that far. This happened in the FCP timeline. I'll play around with it more and see what happened.

Thanks.

1280x720p footage has a Pixel Aspect Ratio of 1.0, while 1440x1080p footage has a PAR of 1.333.
So, if you drop both into a 1440x1080p project, the 720 footage will be distorted. To fix it before you render out, you need to apply a resize to the 720 footage. I think what I would do is drop each clip into VirtualDub, perform the SD resize on each clip, independently, then merge them after resizing.

Thanks Bill. I'll give it a try. So that's what PAR is.

Has anybody found a suitable workflow for Sony Vegas yet?

I am also Vegas users. Have been looking high and low for Vegas workflow too. Tks

For starters, finish your edit on a 1920 x 1080 timeline, then Render as DVDA widescreen template.

Two things I've found helpful so far are

1. Select every clip and set Switches/Reduce Interlace flicker.
2. Make sure render is at Best quality

Results are reasonable but not quite perfect, yet, but other suggestions I've read are to add a fraction (maybe .002) of vertical gaussian blur and maybe a touch of median filter.