GH2 Edit GH2 material directly or transcode with 5DtoRGB?

[Tass]

Hi!

I am going to start to edit a commercial soon that's going to have quite some color correction done to it. I have the free version of 5DtoRGB on my computer, and there is over 40 video files. So it will take some time to transcode all the footage if I choose to do it. I would just like to know if I would gain any quality if I take my time to transcode all the material instead of just editing it in Premiere directly... I did a quick test and I'm no pro at pixel peeping but I did find the transocded material a bit less grainy and blocky when pushed in color correction. But maybe it's just me.

One great thing is that the transcoded footage is much easier for my computer to handle, but other than that, what are your opinions? I have Googled it, some say that they skip it and edit the native footage, other swear by it and would never edit before transcoding their GH2 footage.

I'm running the Orion hack on my GH2 if it matters...

 

[DBP]

Hi!

I am going to start to edit a commercial soon that's going to have quite some color correction done to it. I have the free version of 5DtoRGB on my computer, and there is over 40 video files. So it will take some time to transcode all the footage if I choose to do it. I would just like to know if I would gain any quality if I take my time to transcode all the material instead of just editing it in Premiere directly... I did a quick test and I'm no pro at pixel peeping but I did find the transocded material a bit less grainy and blocky when pushed in color correction. But maybe it's just me.

One great thing is that the transcoded footage is much easier for my computer to handle, but other than that, what are your opinions? I have Googled it, some say that they skip it and edit the native footage, other swear by it and would never edit before transcoding their GH2 footage.

I'm running the Orion hack on my GH2 if it matters...

I've found that my results are similar to yours with 5DtoRGB. I've pixel peeped color corrected clips over top of each other in premiere, and the transcoded clips look a little better. There's some chroma smoothing around edges, and a little less noise over all. There was a comparison show posted Personal View with an red LED lights on a car dash board. The chroma smoothing difference was about as pronounced as I'd ever seen it.

I also like how much more smoothly the transcoded clips edit.

My strategies are to transcode if it's a big project that I can put some time into. If it's just a quick turn around, I don't bother, as the quality difference isn't big enough to warrant the transcoding time.

 

[Tass]

I've found that my results are similar to yours with 5DtoRGB. I've pixel peeped color corrected clips over top of each other in premiere, and the transcoded clips look a little better. There's some chroma smoothing around edges, and a little less noise over all. There was a comparison show posted Personal View with an red LED lights on a car dash board. The chroma smoothing difference was about as pronounced as I'd ever seen it.

I also like how much more smoothly the transcoded clips edit.

My strategies are to transcode if it's a big project that I can put some time into. If it's just a quick turn around, I don't bother, as the quality difference isn't big enough to warrant the transcoding time.

Thanks for your input!

When you transcode, what settings do you use? I use the Apple ProRes 442 HQ setting, 709 matrix (I've heard that "this is the matrix to go with") and luminance range is set to Full (I would really like to know the difference between Full and Broadcast if anyone knows...)

I just found a way to transcode several files at the same time. Rightclick on one of the files, and choose to open it with 5dtoRGB and select it as the standard program for that file type. Now, when you have started to transcode the first file, double click on the next avchd file, choose what you want to transcode it to and start transcoding and so on... I am transcoding 15 files at the same time right now (that would be 15 5dtoRGB windows opened at once)...

 

[M. Gilden]

It depends on what you're doing with the footage. Sometimes it is very beneficial (mostly visual effects work), but other times it really doesn't seem to make enough of a difference.
If you want to get technical, it all really comes down to converting[FONT=helvetica, sans-serif] [/FONT]YCbCr color in the camera to the RGB values that most NLEs use. The editor often does this on the fly, and usually uses a speedy nearest-neighbor algorithm.
In most situations, the end result is almost indiscernible from the original. However, when it comes to things like pulling a chroma key, you may notice jagged artifacts around the borders of your key, or around the bright red LED lights in the examples referenced above. 5D2RGB uses a far better algorithm (lanczos), which converts the color space smoothly and helps avoid the jagged edges.

When pulling a key, I'll throw it into my workflow to help move things along smoothly. Otherwise, I tend to not bother. But that's a personal preference. If you are going to do heavy grades or shoot a lot of contrasty detail (such as illuminated or LED signs) I suppose it might help. I would look at the footage first in the NLE and see if you think it would benefit from it before spending the time transcoding.

Oh, and Tass- depending on the system you are using, sometimes running too many encoders concurrently will slow down the process more than than running them sequentially. Has to do with shared resources and how well your OS can manage tasks.

 

[Admir]

It depends on what you're doing with the footage. Sometimes it is very beneficial (mostly visual effects work), but other times it really doesn't seem to make enough of a difference.
If you want to get technical, it all really comes down to convertingYCbCr color in the camera to the RGB values that most NLEs use. The editor often does this on the fly, and usually uses a speedy nearest-neighbor algorithm.
In most situations, the end result is almost indiscernible from the original. However, when it comes to things like pulling a chroma key, you may notice jagged artifacts around the borders of your key, or around the bright red LED lights in the examples referenced above. 5D2RGB uses a far better algorithm (lanczos), which converts the color space smoothly and helps avoid the jagged edges.

When pulling a key, I'll throw it into my workflow to help move things along smoothly. Otherwise, I tend to not bother. But that's a personal preference. If you are going to do heavy grades or shoot a lot of contrasty detail (such as illuminated or LED signs) I suppose it might help. I would look at the footage first in the NLE and see if you think it would benefit from it before spending the time transcoding.

Oh, and Tass- depending on the system you are using, sometimes running too many encoders concurrently will slow down the process more than than running them sequentially. Has to do with shared resources and how well your OS can manage tasks.

I will try it out and pixel peep a bit more! I am thinking about only encoding the files that are going to be chroma keyed or heavily color corrected... Oh well, horses for courses =)

 

[M. Gilden]

Ideally, it would be great if NLEs started using lanczos by default for converting pixels. I believe that the Mercury Playback Engine in Premiere 5.5 will use lanczos for resizing video if you have CUDA enabled. That means you will get a far better looking HD to SD conversion directly off the timeline for those who currently resort to using external converters for that. I would imagine that it would also use it for converting color space, but I'm not 100% sure. I haven't done much testing with it yet, however since After Effects does not use MPE (or really much GPU acceleration at all yet), if using AE to pull a key this is entirely moot, and converting first with 5D2RGB will help. If color correcting Premiere with a CUDA card, it might be the same thing though.

 

[Disastronaut]

As someone who uses Premiere Pro 5.5 and has a decent computer I've never seen the benefits of using 5DtoRGB, at least in my workflow. I don't shoot a lot of bright red signs or have to pull keys with my footage so the 5DtoRGB transcode is somewhat a waste of time and resources. I do, however, have use 5DtoRGB to get around the limitations of FCP on the rare occasions I have to farm-out stuff to my diehard FCP associates.