Friday, July 13, 2012

PAL to NTSC conversion experiments and results

A tech challenge! I want to enter "Laputa" (7min) into a USA festival which insists on "NTSC" for previews and selected screeners. "Laputa" was shot in "PAL".
In the digital age of hi-def, the issue is running speed. "Laputa" is shot at 25fps. NTSC means a choice of 23.976fps ("24") or 29.97fps ("30").

The "24" looks like the most obvious. "24" is close to "25" aint it? I could go that way by outputting "Laputa" as about 9000 separate still images then re-importing at 23.976. Some fun with audio which would need to be output separately, slowed by 4 percent, get a pitch-up correction and then need some manual sync-up to allow for the 4 percent being an approximation. I had a think about this and decided to go for the "30" because of a more straightforward audio process if maybe more video challenges.

When we set a "25" video editing project to output at "30", this happens by repeating every 5th frame. With "Laputa" I am going "25 Progressive" to "30 Progressive" which cuts out the option of playing some tricks with "Interlacing" to give a smoothing effect. In theory I should expect a "jerky" or "staccato" effect. I am using Adobe Premiere CS5.5 software which offers a "frame blending" option to help out with this. I tried 2 x "30" outputs with and without "frame blending". Both outputs looked good as in pleasant-surprise-good for most of the movie but behaved differently for sequences with a lot of motion. For smooth animated movements the frame-blending worked better. For our fight scenes the non-frame-blend jerkiness seemed to add value while the frame-blend added blurriness to all rapid actions. So like all oldies faced with a high tech challenge I turned to the nursery rhymes of my childhood for inspiration:

Jack Sprat could eat no fat.
His wife could eat no lean.
And so between them both, you see,
They licked the platter clean.

Which means - render out both as high-bitrate intermediates and do an A-B edit with each option covering the motion it handles best. Little detail, I also render the audio separately as uncompressed PCM (max quality) to prevent quality loss with the extra editing stage. I am living with some theoretical video quality loss with this but it looks good to my critical eye.

Some values.
Laputa original settings - 1280x720p25
3 x outputs:
(1) and (2) Intermediates (1)with and (2)without "frame blend". .mp4, codec=h.264, average bitrate = 12 Mbps, peak bitrate = 20 Mbps. VBR 2-pass.
(3) Audio PCM uncompressed 48KHz sample rate 16 bit - ie common "wav file"

New editing project to put (1), (2) and (3) together.
Typical required output is .mp4, h.264, average bitrate = 6, peak bitrate = 10, audio AAC best quality, 320 kbps.