GY-HD101: Any Powerbook editors out there?

[Steve Mullen]

[QUOTE=Matt Daviss]Steve,

The CPU is generating frames based upon I, P, and B frames. Since you can -- I hope -- play HDV smoothly on your PB -- by definition this is a RT process. However, I suspect you mean you can NOT scrub or FF/RR very well. That I believe. I also believe AIC will be much more responsive.

"HDV = the occasional narcolepsy expressed by 15 seconds of beachball for no obvious reason."

That's not good and I can't think of any reason why that should happen. In fact, AIC is far more demanding of the 4200rpm disk!

I just went to the Apple site. They claim: "Do I need a Power Mac G5 to edit HD? No. You can edit HD on any Power Mac G4 with a 1GHz or faster processor and at least 1GB of RAM."

But what is "HD?" DVCPRO HD which is easy? HDV? AIC?

Do you have at least a GB of RAM?

"So, when it's time to output my magnum opus, I guess some shots will be in a position where once was an I frame and now should be a B frame, or whatever. At some point, somebody or something somewhere must (unless I was wrong about the tooth-fairy) regenerate the MPEG2 stream so that all shots and all the transition bits that were created by FCP and now need to also be part of the MPEG2 stream, must follow the strict IBBPBXYS-whatever pattern from frame 0 to the bitter end."

There's a lot in that question!

1) With HDV, according to Apple, when there are no FX the source HDV file is copied to the output file. Nothing is "conformed."

2) With HDV, according to Apple, at each cut-point ONLY the frames at the cut-point are decoded and recoded.

3) With HDV, according to Apple, when there are FX the source HDV file decoded and recoded to the output file. This is "conformed."

4) With AIC, according to Apple, when there are no FX the source AIC file is must be decoded and recoded to HDV and sent to the output file. Everything is "conformed."

5) With AIC, according to Apple, when there are FX the source AIC file decoded and recoded to the HDV output file. Everything is "conformed.

Clearly, a cuts-only HDV timeline should export to HDV tape with far, far less delay that were it an AIC timeline. Do you not find this to be true? When there are FX, there should be little if any difference between AIC and HDV.

Likewise, if you export to any other format, there should be little if any difference between AIC and HDV.

Which brings-up to issues:

1) HDV may require dual CPUs and far more than 1GB to really work well. Apple seems very clear, privately, that the minimum HDV machine is a dual 2.0GHz G5. Using anything less, seems unacceptable to Apple. So unaccetable they replaced my loaner dual 2.5Ghz with a dual 2.7GHz!

2) According to Apple, when I pre-reviewed FCP 5, only the 2 frames at each cut point need to be processed. That's how native HDV is supposed to work. They say it should feel like DV. BUT, for some reason, FCP goes off on a pilgramage. In short, it seems to be doing something that you call conforming. I call it BS because: if FCP works as Apple says it does, then it is doing something else; whereas if it "conforming" then FCP doesn't work as Apple claims, hence we have PR BS. How do we get Apple to fess-up -- I don't know.

Now about AIC. My iMac G5 1GHz required 4 hours to capture 1 hour of 720p30 and transcode to AIC! Are you able to capture in RT to AIC?

Hopefully, this discussion will clarify what FCP really requires to work with HDV.

[Matt Davis]

Have performed a couple of tests on my 1.33 GHz PowerBook 17" with 1 GB RAM - not totally scientific, but as good as I can get in the limited time I get between work and first-born's bathtime:

Created a 1 minute sequence of 3 HDV clips on an HDV timeline.

Exported sequence as HDV without anything done to clips other than two cuts between the three clips. It took 2 mins, 16 secs to export the 1 minute sequence (it reported 50 mins at first, but that dropped dramatically as the export progressed).

Hooray, but nothing like real world - I am sure you advocate shooting HDV flat and using post to bring out colour and contrast in a wider colour space before export to final delivery format.

I scrubbed each clip so different frames were at the I and O points, then added colour correction to all three clips and the second clip's scale was reduced to 75%. I exported the clips again, same settings, and FCP estimated 14 minutes, and finally completed at 15 minutes 4 seconds.

I scrubbed the clips again so any pre-rendered clips were null and void. I then rendered all the green-bar bits (the whole thing - and it took a little over 15 minutes - funny, that).

On opening the render files that FCP made for this project, all were in HDV1080i50 mode.

Thus, my empirical evidence supports my original hypothesis that in HDV projects, FCP may render effects in 4:4:4 (thanks for clarifiying that) but stores the results in the compressed 4:2:0 HDV format at 19 Mbits per second. I would prefer the results of my effects and colour correction (rendered at 4:4:4) to be stored in a 4:2:2 colour space for final delivery in my chosen format rather than recompressed to the significantly limiting HDV format.

Hence, I will shoot in HDV flat in a 4:2:0 colour space, edit in a codec that stores 4:4:4 renders in AT LEAST 4:2:2 colour space (maybe AIC, maybe DVCPRO-HD with a few tweaks to rez) because I don't consider HDV to be an editing format, simply a convinient acquisition format.

And a PowerBook 17" at 1.33 GHz with 1 GB RAM (here we haul ourselves back on topic) can edit HDV even with filters and effects, but output is slow slow slow. For my corporate stuff with sensible time scales, I'll continue to shoot HDV and edit in a wider colour space to deliver in a file format such as QT, WM9 or AVI. For long form, perhaps we can downconvert with offline just like SD. For quick turnaround, though, I am sticking to SD!

[Steve Mullen]

[QUOTE=Matt Daviss]Have performed a couple of tests on my 1.33 GHz PowerBook 17" with 1 GB RAM.

Created a 1 minute sequence of 3 HDV clips on an HDV timeline. It took 2 mins, 16 secs to export the 1 minute sequence.

Sony or JVC?

So a cut-only is about 2:1 -- which seems like a long time to decode & recode a few frames at the cut-points -- unless you were working with 1080i which I found out must conform every frame.

If you were working with 720p30 this is where what Apple claims happens with HDV differs from what folks find. It should be a copy from source to destiation file with 720p.

"I added colour correction to all three clips and the second clip's scale was reduced to 75%. I exported the clips again, same settings, and FCP estimated 14 minutes, and finally completed at 15 minutes 4 seconds.

This makes some sense because you are rendering all the FX before export even starts. Try rendering to uncompressed.

"Thus, my empirical evidence supports my original hypothesis that in HDV projects, FCP may render effects in 4:4:4 (thanks for clarifiying that) but stores the results in the compressed 4:2:0 HDV format at 19 Mbits per second. I would prefer the results of my effects and colour correction (rendered at 4:4:4) to be stored in a 4:2:2 colour space for final delivery in my chosen format rather than recompressed to the significantly limiting HDV format."

AH -- but here's the deal. Yes it stores the renders as HDV as you expected. But that's just so you are able to play complex segments fast in the Timeline.

BUT, when you EXPORT to something other than HDV -- it starts fresh with all HDV sources & decodes & renders the FX all as 4:4:4 uncompressed. Only at this point does it compress or encode to the format you want. So with FCP you always get exactly one decode of MPEG-2 & one compress or one encode. FCP is able to not to resuse the render files unless you want it to in order to save time.