Editing H.264 files directly seems to... work?
by Jarkko Laiho
on
Jul 6, 2009 at 6:24:49 pm
Here's something I don't understand.
I just got a neat Sanyo Xacti VPC-HD2000 camcorder today. It shoots up to 1920x1080@60fps, saving MP4 files (H.264/MPEG-4 AVC).
The full resolution is too much for my current hardware to handle (Core Duo iMac), so for the moment I'm shooting 720p@30fps.
Having read up on the topic previously, like this thread here, it seems to be an accepted fact that H.264 isn't an editing format, which is logical, given that it's a delivery format by nature.
Why, then, does FCP 6 seem to handle the raw files from the camcorder just fine, without any rendering that I can see? (QT version: 7.6.2)
To test this better, I created a new, empty scratch disk and a new FCP project with a single sequence. The default sequence was set to PAL DV, which I've been using previously. I imported a one-minute 720p MP4 file from the camera and dropped it on the sequence so it would automatically change to the correct settings. After it did that, the (relevant) sequence settings are as follows:
General tab:
Frame size: 1280x720 -- Preset: HDTV 720p (16:9)
Pixel Aspect Ratio: Square
Field Dominance: None
Editing Timebase: 29.97
Compressor: H.264 -- Quality: 100%
Audio Settings: 48kHz, 16-bit, Channel Grouped
Video Processing tab:
Always render in RGB selected, options grayed out (apparently due to H.264 only supporting RGB)
Timeline Options tab:
Starting Timecode: 01:00:00;00, Drop Frame checked
Render Control tab:
Codec: Same as Sequence Codec
What's odd to me is that the render status bar is grey for video, indicating no need to render it. The tooltip says: "Video: Media File. Audio: Needs Render. RT Extreme does not support this sequence".
I can edit the clip completely normally, as far as I can tell. Rendering All for a 54-second clip takes a few seconds to render the audio, but that's it. The video isn't apparently processed at all. The generated audio render file takes 20 megs, but only five files totaling about 600 kilobytes appear under Capture Scratch/Render Files/Constant Frames. I even created a simple Motion project, did an effect and imported that back to FCP just fine.
I was under the impression that to edit anything produced by the camera, I'd have to export the clips to e.g. ProRes 422 first. Apparently not.
If I understand correctly, when I work in Motion, the "Same as Sequence Codec" setting in Render Control means that after I import my finished Motion project to FCP and render, the Motion effects video track is encoded with H.264 in accordance with the sequence setting. This sounds risky; having V2 be encoded separately from V1 with a not-insignificantly lossy codec sounds like a one-way trip to artifact city to me.
The question is: what is the suggested workflow in this case? At which points do I need to convert all or some of the footage to ProRes or something else? The aim is naturally to minimize the amount of quality loss in the course of working with the footage in FCP, Motion and SP.
Uninformed guess: do the raw edit in H.264 for simplicity, then convert just the necessary, edited-in bits of video to ProRes for Motion work etc. Advice, pointers to relevant manual bits or website tutorials on how to do this are more than welcome.
Bonus newbie question: is the very process of rendering in FCP lossy, if the sequence preset uses a lossy codec (including ProRes)? Do I need to worry about changes that require rendering to my source video track, i.e. does each render mean a re-encode and an inevitable quality drop?
Re: Editing H.264 files directly seems to... work? by Jeremy Garchow on Jul 8, 2009 at 12:44:50 pm
Well, you can edit it, but make a cut and put a dissolve (or something) between clips and watch what happens. You will also not get real time output of any video capture card. What you have told FCP is to make an h264 sequence so it won't need rendering, but h264 is a web delivery codec.
[Jarkko Laiho]"The question is: what is the suggested workflow in this case? At which points do I need to convert all or some of the footage to ProRes or something else? The aim is naturally to minimize the amount of quality loss in the course of working with the footage in FCP, Motion and SP"
Do the transcode before you start editing. Take the raw files to Compressor and make ProRes files which match your footage and transcode the audio to 48k16 or 24 bit audio. This will be much easier to use during the course of the edit and ProRes is a video codec, not like H264 which is a lossy web codec.
Re: Editing H.264 files directly seems to... work? by Jarkko Laiho on Jul 8, 2009 at 1:23:46 pm
"Do the transcode before you start editing. Take the raw files to Compressor and make ProRes files which match your footage and transcode the audio to 48k16 or 24 bit audio. This will be much easier to use during the course of the edit and ProRes is a video codec, not like H264 which is a lossy web codec."
