The Transcoding experiment begins...

Shooter

Shooter

Veteran
I realise this is probably more suited to the POST FORUM but I also think it is something that others here are now driven to thinking about due to the emergence of this AF100 camera. (AVCHD) .

Maybe not such a big deal for FCP users who are used to it and have their PRORES intermediate workflows in place already.

Despite PPRO CS5 editing AVCHD natively... it is still a long GOP Codec and I wanted to know where to fall back to if I got grief at any stage with it. I know I can edit it natively and have briefly tested that. It certainly does not work quite the same with effects , CC and some plugins etc..but I need to test and identify these better.

SO....I really wanted to test transcodes to an intra farme codec...and I used Adobe Media Encoder (Stand Alone) to directly transcode the MTS source files. I imported the results ( new files) into PPRO and opened them in their own matching sequences. I viewed everything on a calibrated HD monitor (HP Dream COlor) direct from the timeline and could swith between sequences (on the same frame)

so far: (Step 1)

1> AVCHD to DVCPROHD 720p
2> AVCHD to DVCPRO HD 1080
3> AVCHD to AVC Intra 720p
4> AVCHD to AVC Intra 1280

All edited like a dream and all looked absolutely fine and comparable to original and each other.

Absolutely no obvious "loss" of anything res or color. Have not run the clips thru the scopes yet.

Obviously the AVC Intra codec created very large files.

Main thing I learnt. Transcoding is a pain due to loss of time and efficiency.

Shoot / ingest / edit compared to Shoot / ingest / transcode /edit


Next Step will be test other Codecs eg AVID DVXn


BTW: I already shoot as if I am shooting 35mmm film ( economical camera run / cut)
 
I don't see how you're not seeing an obvious loss of resolution when transcoding to DVCProHD 1080 or 720 or AVC Intra 720. You're actually losing resolution with those. There's dramatically less pixels so you should see an obvious loss in resolution unless you're watching it in a very small window on your computer. Try watching those full screen or outputting to an HD TV and you should see a difference.

Also, you said it yourself that you need to transcode anyways for some effects, filters, and color correction, so why not just transcode at the beginning and get it over with so you don't have to worry about it later? As far as time for transcoding you can do it as a batch process. Start is and walk away until it's done. I admit that it would be nice to edit natively, but the more I look at it and realize that you'd still have to transcode at some point so native editing isn't really a big deal.
 
I would try out either DNxHD or Cineform...I don't know alot about DVCProHD, but I remember hearing it was not a good idea to convert to those codecs (99.9% of the time I work native, in an extreme case I go DNxHD)
 
Try watching those full screen or outputting to an HD TV and you should see a difference.
Click to expand...

I would have thought an HP Dream Color HD monitor ( full screen) would have adequately displayed a result worth considering.

Recommend another monitor (or the one you use) and I will try and see what you see.

BTW: The 720 tests were just as a comparative reference for the pixel to pixel transcode tests I also ran. My workflow only eventually needs to consider that my TVCs when broadcast are PAL 720 x 576.. Even when I shoot 35mm film or Red 4k... they are eventually posted as... then broadcast PAL 720 x 576 FHA

What are the broadcast specs of your projects in your region?

(Interesting that the high end Pannys use AVC Intra and DVC Pro HD)
 
Hmmm... they broadcast pretty low resolution where you are it seems, although it's higher resolution than the old SD in the US (480p). Here in the US, HD channels are 1280 x 720p or 1920 x 1080i broadcast. It depends on the channel.

I don't know why you aren't seeing a difference between 720p and 1080p or the AVCHD and the DVCProHD on your monitor. I hear those HP Dream displays are pretty good, but I've never seen one in person. I can see a difference between different resolutions of the same clips on my MacBook Pro's 17" LCD pretty easily (native resolution of 1920x1200), as well as when I output to my 60" 1080p plasma via the MiniDisplay Port or my Matrox MXO2 Mini.

AVC Intra is a good codec and is full HD 1920 x 1080p. DVCProHD on the other hand I couldn't explain why Panasonic is still using it. Maybe because there's infrastructure already in place out there for it and they don't want to abandon companies that are still utilizing that infrastructure.
 
Premiere Pro CS5 can also transcode to XDCAM HD422 in either 8 or 10-bits. At a constant 50Mbps bitrate, these files are half the size of 100Mbps AVC-Intra videos. The AVCHD source videos are 24Mbps @ 4:2:0 8-bit, and to my eyes, XDCAM HD422 does an excellent job of preserving color depth while maintaining reasonably compact files.
 
DVCproHD hd is 1280x1080 and 960x720 for 1080 and 720 respectively. They are not full raster and you are loosing resolution when you convert to this codec. If that matters to you or doesn't is your call, but you are loosing resolution, no two ways around that.

Panasonic uses it because it is a decade old codec, and for a long time there was nothing better. If you notice their new flagships are based around AVCintra as the future codec of choice.
 
Last edited:
I'd say if you can avoid transcoding for ingest, avoid it, and if you cant then if you can avoid DVCPROHD, avoid that. I hardly see any possible reason to transcode for a CS5 workflow. Why do you feel you may need to transcode? Are you not getting realtime playback? I get stable realtime AVCHD/h264 playback on a budget 4 year old laptop, so its hard to imagine you would have problems unless you found a way to run Premiere CS5 on a pentium3. Fewer generations will give you better quality and save you a lot of time.
 
nyvz said:
Why do you feel you may need to transcode?
Click to expand...
If you can manage all your comps in Premiere and do a single render to your distribution format, then there is no need to transcode. If you need to do any intermediate renders, however, writing back to AVCHD format will degrade your render to its 4:2:0 color depth. That's what makes 4:2:2 transcoding a better choice. For this, you need ProRes, 100Mbps AVC-Intra, or (my preference) 50Mbps XDCAM HD422.
 
Last edited:
lpowell said:
If you can manage all your comps in Premiere and do a single render to your distribution format, then there is no need to transcode. If you need to do any intermediate renders, however, writing back to AVCHD format will degrade your render to its 4:2:0 color depth. That's what makes 4:2:2 transcoding a better choice. For this, you need ProRes, 100Mbps AVC-Intra, or (my preference) 50Mbps XDCAM HD422.
Click to expand...

I certainly will be testing other codecs as well. Thank you for the suggestions. ( Its all a new learning curve for me with the aim of best results and best workflow.)

BTW Transcoding to 4:2:2 does not make it 4:2:2. It only pads the data I believe.

BTW : HP Dream Color for your reference http://h71028.www7.hp.com/Hpsub/cache/596803-0-0-225-121.html

As regards reasons for transcoding....if we back the bus up a little here....in this forum it has been stated "AVCHD is an acquisiton Codec, not an editing or delivery codec."

Has anyone actually tested AVCHD (ex AF100) with their usual plugins, transitions and effects from their preffered NLE.? eg Ultra Key, Key Light, Twixtor, Colorista II, SA, Magic Bullet Looks, Time remapping, Blend Modes etc....

Are you certain they all work at least as well as your current edit / composite / CC experience ?
 
Shooter said:
BTW Transcoding to 4:2:2 does not make it 4:2:2. It only pads the data I believe.

Has anyone actually tested AVCHD with their usual plugins, transitions and effects from their preffered NLE.? eg... Colorista II
Click to expand...
Yes, transcoding from AVCHD 4:2:0 to a 4:2:2 codec won't increase the color depth of the original source file. However, the transcoder will generate a fresh stream of encoded data from the decoded RGB frames generated by Adobe Media Encoder. It's important to save the newly encoded data in 4:2:2 format to preserve the full accuracy of the transcoded frames.

I've used Colorista II in Premiere CS5 to grade AVCHD source files shot by the GH1 and it's worked well for me. The major compromise I've found with all non-RAW formats is that there's no accurate way to adjust the color temperature that's burned into the encoded file. For that reason, I feel it's best to fine tune the color temp and film mode in-camera as close as possible to your final desired look. Since the camera has direct internal access to the 12-bit RAW sensor data, it can do a more precise color correction than any NLE is able to do in post.
 
I have done plenty of color grading tests using Color Finesse (AfterFX CS5) and Looks (Premiere CS5) with the native and DNxHD converted AF100 footage...very little difference was found. However you do get some weird "jitter" and artifacts during playback if you have not rendered in premiere, hopefully this will be fixed on an update soon.
 
lpowell, perhaps you can tell me, I'm experimenting with your advice to transcode AVCHD to XDCam 422 50 and I not only have to wait for the transcode, but then after importing the transcoded footage, I have to wait for 8 tracks of audio to conform, even though there are only two in my footage, and the other 6 tracks are no faster. How do I set it to create only the two tracks from the original file?

Many thanks
Lucas

BTW: The footage certainly looks good, but this conforming is just driving me insane. It seems to never stop, and while it's crunching numbers, my video is as choppy as hell.
 
Thank you for confirming ( as below)

So we now get to the stage that the transcode is sorted and the benefit is a new stream of Intra frame footage. The disadvantage remains. We have used more than twice as much disk space by using an intermediate.

I am thinking that I will now transcode as I Ingest from my external drive to my local (NLE) drive(s).

Previously I just ingested ( transferred) direct from external drive to local drive. eg P2 Card to external, external to local. External became permanent Backup /archive of source file.

I think this is going to work out fine.

lpowell said:
Yes, transcoding from AVCHD 4:2:0 to a 4:2:2 codec won't increase the color depth of the original source file. However, the transcoder will generate a fresh stream of encoded data from the decoded RGB frames generated by Adobe Media Encoder. It's important to save the newly encoded data in 4:2:2 format to preserve the full accuracy of the transcoded frames.

I've used Colorista II in Premiere CS5 to grade AVCHD source files shot by the GH1 and it's worked well for me. The major compromise I've found with all non-RAW formats is that there's no accurate way to adjust the color temperature that's burned into the encoded file. For that reason, I feel it's best to fine tune the color temp and film mode in-camera as close as possible to your final desired look. Since the camera has direct internal access to the 12-bit RAW sensor data, it can do a more precise color correction than any NLE is able to do in post.
Click to expand...
 
lpowell said:
If you can manage all your comps in Premiere and do a single render to your distribution format, then there is no need to transcode. If you need to do any intermediate renders, however, writing back to AVCHD format will degrade your render to its 4:2:0 color depth. That's what makes 4:2:2 transcoding a better choice. For this, you need ProRes, 100Mbps AVC-Intra, or (my preference) 50Mbps XDCAM HD422.
Click to expand...

Ah ok, why would one need to do an intermediate render? Only thing I can think of is for VFX shots from another application or maybe too many complicated effects to be handled by the GPU (or not supported by GPU) in realtime otherwise. I havent had any such situation occur yet, between the CPU and GPU processing available in CS5.
 
You must log in or register to reply here.
Back
Top