XHA1 & Premiere quality issues. NEED HELP!

[PDR]

It was an old behaviour of older Premiere versions where low quality render previews were used for the export instead of the original files. I just do it now out of principle, I don't know if it still behaves that way. The option in the fly out menu in AME is still there, however. So double check that before your final export

You can automatically re-link the actual files when you swap for final render. Here is a related guide
http://www.dvinfo.net/forum/canon-eos-5d-mk-ii-hd/144428-premiere-pro-cs4-solution.html

 

[Kevin Dorsey]

HD footage shot 24f from an XHA1 has no pulldown. It's recorded at 24. No pulldown removal is needed. Captured files from HDVsplit will read out at 23.976 and will be immediately editable in PPCS3 and PPCS4.

It's possible the quality issues are a result of how Premiere displays the footage in the program monitors, and are not an accurate display of how the final project will look once exported in the proper manner. I would do a short test of both Final Cut and Premiere and check the final output to see if one has a quality advantage. I'd be very interested in seeing those results.

Premiere does a poor job of downconverting HD footage. It looks very soft. I get much better results using TMPGEnc. I have Cineform on my system, so I always export an HD Cineform master, then import that into TMPGEnc for encoding to DVD or whatever my final delivery is. In the past I've downconverted using After Effects and the results were very good. But I had some audio sync issues.

If you want to send my your Mac I'll do a side by side test.

-Kevin

 

[PDR]

HD footage shot 24f from an XHA1 has no pulldown. It's recorded at 24. No pulldown removal is needed. Captured files from HDVsplit will read out at 23.976 and will be immediately editable in PPCS3 and PPCS4.

That's what I had read too. But I am 100% certain the sample he posted exhibits 2:3 pulldown, with several field repeats at the beginning. It might be that RPK did something weird during capture. He mentioned that FCP had no problems, and I tested on vegas 9 and it had no problems either.

I verified the problems on that sample with PP CS4 as well. They are real, at least with that sample.

Premiere does a poor job of downconverting HD footage. It looks very soft.

Agreed. Premiere does have a poor downscaling algorithm. Premiere also has a very poor deinterlacing algorithm. With avisynth , you have much better control & quality

 

[RPK]

Captured files from HDVsplit will read out at 23.976 and will be immediately editable in PPCS3 and PPCS4.

Yes, they do and they are, but once the footage is in Premiere I get the artifacts that can be seen in the stills from my first post. Same thing when I capture directly in Premiere.

It's possible the quality issues are a result of how Premiere displays the footage in the program monitors, and are not an accurate display of how the final project will look once exported in the proper manner.

What you see in the samples is exactly how my footage looks when exported, either via AME or directly exported and either video or stills.

Premiere does a poor job of downconverting HD footage

I have these issues at full 1080p without any downconverting.

If you want to send my your Mac I'll do a side by side test.

I get the same artifacts when using Premiere CS3 or CS4 on PC or MAC with various ways of capturing with XHA1 or HV20.

It might be that RPK did something weird during capture.

I was hoping that it's a simple to fix capture issue, but capturing with HDVSplit is a no brainer and there doesn't really seem to be much that can be done wrong at that stage.

But I am 100% certain the sample he posted exhibits 2:3 pulldown, with several field repeats at the beginning.

The frameserver workflow you posted gives perfect results in Premiere with my footage! Still have to capture that 45 minute long clip for testing audio sync.

RPK

 

[RPK]

I would do a short test of both Final Cut and Premiere and check the final output to see if one has a quality advantage. I'd be very interested in seeing those results.

Yep, that's the comparison you see in the sample stills in my first post!

 

[Kevin Dorsey]

I'm glad you're getting it worked out. I'm still a bit confused as to why there is any pulldown removal needed. The XHA1 only has pulldown if it's shooting in SD 24F, or if the camera is set to downconvert in camera HD 24F to SD 24F. The second option is where there are no pulldown flags. The HV20 on the other hand, shoots 24f inside a 60i wrapper, and there are no pulldown flags. Is it possible the footage was shot on the HV20?

 

[RPK]

I had some time over the weekend to try the frameserver workflow with a long XHA1 clip. The footage was from a 50 minute long musical performance, and I added and manually synced up externally recorded audio.

Unfortunately the externally recorded audio does drift out of sync with the footage very noticeably, especially noticeable in the percussion. Comparing the frameserved footage with the original clip, the drift seems to be 2 to 4 frames at around 35 minutes into the footage.

I will try if using the DGindex workflow remedies the sync issue but haven't had the time for that yet. I guess I will also check out the trial of Cineform's NeoScene.

A mistake I made describing my problem at the very beginning of this thread: It seems that captured clips I add to the timeline do in fact result in a red render bar!

I have to check a few older clips to see if those didn't cause a red render bar, but this long clip definitely does not conform to the HDV 1080p preset straight out of Premiere capture.