Recently there have been two threads on the Digidesign User Conference on this subject.

One asking about the best codec to put the least strain on the host CPU

usastra asks

I have a session for a live concert with a quicktime movie using the H264 codec. When the movie is online it crashes very often, especially if I set the cursor to a new place in the timeline.
If the movie is offline PT is way more stable (though still with the occasional crash)..
I was wondering if the session could be made more stable if I render the movie with a different codec and which codec would be best so I would have the least CPU strain?

The other commenting on the load playing an mp4 file has on the host CPU.

Roger Stauss asked…

I’ve been using an Mp4 vid to ease the up/download time. The PT session runs with some latency. It’s very smooth with a QT DV. Is there a way to smooth out the session operation? (I run Video on separate drive)

I have compiled the best responses from both threads below…

Chief Technician advised simply…

Use the DV codec.

c.evans agreed…

Agreed. The DV codec works well on my system, but what really helped is adding the Canopus ADVC 110 and sending my video out to an external monitor. Also, it is very important that your video is playing from a separate drive than your audio. I even use the internal SATA buss for my video when using firewire for video. It also helps to have a third party PCI firewire card to add some beef and options to your system.

Craig F chipped in…

mp4 is a intraframe codec same as h264 hence heavy CPU usage.

José Luis Díaz explained…

Always choose intraframe codecs (DV, Motion JPEG-A, Photo JPEG, etc).
Always avoid interframe codecs (H264, Sorenson, etc).
Interframes codecs demand much more brute force from CPU than intraframe codecs.
Intraframe codecs only compress the data of one frame regardless any adjacent frame.
But to decode a simple interframe stream of frames (like H264) the CPU of your computer will dedicate a lot of its cycles just for that simple moving image because it must to compute differences and similarities between frames.
Check all this with Activity Monitor App.
Play a PT session with an interframe movie. Them the same with a intraframe movie.

Jon_Atkinson added….

Digidesign Technical Support have gone on record here a few times saying that H264 has been the root of many PT related QT problems. “A can of worms” was, I believe, the phrase used….

usastra responded…

Thanks for all the replies! For now, I have converted the videos to mjpegB and the session runs much smoother! I will experiment with other codec later.

philip888 asked….

What are you using to convert with?

usastra replied….

Quicktime Pro

Rick Sanchez confirmed….

A couple of solutions I found that worked well for me.

1. Once you’ve downloaded the video, use Quicktime Pro to transcode it to DV. If you have a fast computer, it’s pretty quick to do. Usually much quicker than the download time.

2. Tell the editors to send you Motion JPEG A video instead of MP4/H.264. It’s still very small and plays well with ProTools.

One of the falsehoods about H.264 is that it is not a frame accurate format. This is not really the case. It can have a variable frame rate, but that is a setting when you do the original encode/capture. What it is, is a long GOP (Group of Pictures) format. Much like Mpeg2. In the first frame, all of the pixels are captured. In between the first (key) frame the next (user selectable) key frame, the computer looks for differences between the the Key frame and the next frame and only encodes anything that has changed. If the pixels (say a non-moving background) stay the same from frame to frame, it just fills in the blank with already captured pixels. Then when the next Keyframe comes around again (I think the default is 24) it captures all of the pixels and starts over. It’s much like creating an audio ambiance fill track for a scene. You only need a few seconds of fill and then you loop it until there is the next change in ambiance.

I take no credit to the advice I have quoted here but do confirm it from my own experience.