mindlessv

The talk hairy_hen linked is a great series of explanations for audio lay people. Fun that the speaker uses Decibel for playback; that's my software of choice for bit perfect audio.

One distinction I'm not sure the speaker makes directly but can be inferred in places, is that there is a difference between double blind A/B with inexperienced listeners of unknown content and experienced listening of intimately known audio, the latter of which is inherently more biased but is actually a better metric for music (really anything that's enjoyed repeatedly). Engineers like myself are generally taught that the more unbiased the testing method the better, and yet a truly double blind test in the former case is not fully capable of testing many things that matter, such as emotional engagement and long-term enjoyment (in gamer-speak, replay value).

Put another way, one may notice but not mind if some random film they watched once didn't look that amazing, but don't you dare mess around with Star Wars! ;) Watch the random film a few more times and the nuances and flaws become more noticeable each time. It's true for the plot and storyline, it's true for visual quality, and believe me, it's true for the audio quality.

All that to say, I have some good analog gear and the Despecialized Edition sounds pretty great, from the first time to the tenth time, and onward. hairy_hen did a brilliant job.

Just to be clear, you've tried position-only stabilization (no subspace warp), correct? Seems like that would be excellent for gate weave once tuned a bit.

Warp Stabilizer effect in AE is a pretty decent stabilizer. It can even do content aware edge recreation to avoid cropping.

And yet I do, as is evident in the picture I posted.

Broken link for me. :(

Premiere has a better Interpret Footage dialog. You can create and name a new 23.976 sequence, import the DV file, interpret as 23.976 and progressive, and drag the file to your sequence. Then save project and import the project into AE. The resulting sequence(s) will play back correctly.

If you're still getting combing after this, then it's baked into the source. Try upper/lower field. If neither works or only partially works, it's maybe telecine; in AE's Interpret Footage try Guess 24Pa Pulldown.

Edit: clarity.

I assumed that setting the fields to none would set the video as progressive.

Comp/project settings no, this would not affect your sources. Only the timeline and how the output is marked.

If you Interpret Footage > Separate Fields: None then you should have no combing, assuming the source footage is indeed progressive.

But my AE Composition is set to 23.97fps, so I don't see any reason, why interlacing artifacts should appear, if the video is set as progressive.

How are you setting the video as progressive?

Chewtobacca said:

It won't.  I understand what you are saying about not wanting anything to be altered at this stage apart from the color, and that's fine, but you shouldn't apply RFF flags to 29.97fps video.  At some stage, you will have to slow it back down to 23.96fps and then encode with RFF flags i.e. apply pulldown.

This. ^

Ideally, what I would do in this case is use another tool such as AviSynth to interpret the footage as 24000/1001 progressive and save a lossless master. IIRC, DV/DVCPRO supports neither of these things, so it's at best suitable as an intermediary.

Puggo: This makes more sense. Assuming the cam was used more like a still camera, taking one image for each frame, then there should be no combing. In which case, Chewtobacca is absolutely correct that the footage should be interpreted as 23.976 (24000/1001).

Unfortunately, AE's interpret footage dialog doesn't seem to play nice. I see no option to assume a different input rate, only to conform to a new rate. So you'll probably want to run the source through AviSynth with the AssumeFPS() directive and use the resulting file as your source.

To be clear, you likely have 23.976 fps progressive footage, marked as 29.97 fps and marked as interlaced. You want to interpret it as 23.976 fps and non-interlaced. The latter can be achieved in AE by setting separate fields: none, however there seems to be no easy way to fix the input fps in AE.

Something tells me that the capturing device itself was operating in full interlaced mode, since that's typical for consumer devices of SD vintage and is indicated by DVCPRO. I highly doubt the capturing device converted film 23.976 to video 29.97 via telecine in this case. Either progressive marked as interlaced or full interlaced video.

If you can send me the sample I'll be glad to lend another set of eyes (and tools). Used to do this for a living.

Harmy, your part is excellent, however a bit long. I think the audio works well at that length; I suggest cutting to other video one or more times while you're talking, which may break things up a bit and depending on content, provide more visual information.

If I lived closer I'd shoot some B roll of you walking down the sidewalk and you could rotoscope a lightsaber (color of your choosing)...

Barring that, an over-the-shoulder shot of you editing would be excellent. ;)

When you get a new HDD, buy two and send one loaded with backup files to someone you trust. ;)

Mac users can use this little GUI with the xdelta patch: http://projects.sappharad.com/tools/multipatch.html

I verified the resulting md5 matches.

jzilli said:

Harmy said:

Yeah, it's supposed to be the TB 97SE.

Also, here's a completed comparison gallery:

https://plus.google.com/u/1/photos/109609428403596349302/albums/6044645069501365185

The time, patience, attention to detail, quality of work, and thoroughness..... BRAVO

 This! ^^

I'm not certain the max bitrate the PS3 supports, but if you can't remux then I suggest using x264's constant quality mode. RF 16-18 (lower is higher quality) and tune "grain" to preserve the film grain.

Normally I'd suggest a higher RF number but typically more quality is needed when recompressing multiple times. x264 does a relatively good job in these cases but a slightly lower RF (more bits) makes the conversion more transparent.

What issues is handbrake having?

The whole answer is far too off-topic but the gist of it is the extremely long GOP with few I-frames and SPS/PPS requires a workaround.

If anyone has any further HandBrake related issues feel free to PM me; let's keep this thread on track.

And of course, many thanks to Harmy! Brilliant work. Looking forward to the high bitrate MKV.

[Minor edit for clarity.]

davextreme said:

Subler, which I use for messing with video metadata, recommends changing the H.264 Video Profile/Level setting to "Main @ 3.1" for Apple devices. The Star Wars mkv came in at High @ 4.1. Do you think changing this will fix it, or should I plan to recompress with Handbrake.

Subler's great but it doesn't recommend changing profile/level, only for a specific, unrelated issue. Feel free to PM me if you need any further help on this, kind of off topic.

Harmy said:

Also, I was wondering, would there be any technical disadvantage to making the container of the smaller cropped version mp4 instead of mkv?

See my last. While MP4 certainly isn't limited to Apple devices, they do not play well with sliced video. For maximum compatibility I'd recommend disabling slices for an MP4. They're only needed for Blu-ray compatibility anyway.

Edited to add: Also, MP4 is more limited in the subtitle formats it supports and DTS isn't a thing.

davextreme said:

I'm going to run the film through Handbrake to watch on Apple TV.

Infinity said:

First off don't use Handbrake as that will compress the video yet another time (and possibly the audio depending on your settings). The best quality for your purposes would be from remuxing the video from the large MKV release and the DD 5.1 audio from the AVCHD release.

Unfortunately, AppleTV doesn't play well with h.264 sliced coding. It will play the video but may skip and heaven help you if you try to seek. While I fully agree that remuxing is generally a better choice, recompression is unavoidable in this case.

Don't rush to convert the v2.0 AVCHD, though; HandBrake trips over the video stream but we're looking into why.

...you might find someone here that's better suited for voiceover work.

Funny you mention that...

Earlier today I was watching the first sources doc thinking Harmy's voice is actually good, but the recording could use a proper mix and mastering session. So I remastered the whole and it sounds much better, despite starting from a compressed source. Of course it would be even better if I had separate tracks for voice and background music.

I certainly don't want to step on Hairy Hen's toes, or anyone else's for that matter, but if nobody has stepped up to offer audio engineering for the documentaries I'll be glad to lend my talents.

Great job! Congrats on ESB DeEd v2.0 also!

The inline image and linked gallery in the first post are broken. Is there artwork for this or were they all samples?

Yes, a hash is a hash, no concern here.

Indeed it is.

My checksums match what Harmy has stated, and the byte size is 19,213,754,660 as obtained via Nearsighted Scrappile's distribution. So, someone is mistaken.

Harmy, if you would please double verify the byte count and/or hash(es) of the corrected version.

It's not impossible to purposefully break MD5 these days, but by chance? Pretty impossible (never actually happens, ever).

I see no file size difference between the MKV and what was previously stated, anyway.

Nearsighted Scrappile said:

Chewtobacca said:

I think he's more concerned that he's been providing people with the unfixed version.

Nearsighted Scrappile, the myspleen version is fine, and there are no duplicates.  There were checksums posted a few pages back.

Thanks mate and yes you are right about my concern. I'll make sure to get the corrected one shared and the old one nixed once I get it from the Spleen. I find it ironic that someone who likes a band called The Strokes asked me if I freak out much, LOL.

I checked the MD5 on your work here and it matches. No glitch that I can perceive at the specified time.