[Psycogeek]

Ok i am going to try and doccument what I do (myself) to convert a video
that is already compressed for other reasons,
try to get it within the limitations of ONE item the MP4 Watch with a 1.5" screen 128x128 video and 1gig of memory.

First Peel off the AUDIO using goldwave (free to try)
Save it out as a 44.1k wave file like a standard cd file
WHY
MP3 and other various audios just dont seem to stay in sync if you try and compress and recompress them again and again,
also any frame errors in the video will take the whole thing out of sync,
by removing the audio it can better match the video and a preview of the combination is more reliable.

In Goldwave (or other audio program) this is your oppertunity to clean up the audio.
do a little equalising or low pass filtering to hack off the highs
by killing all the audio above 4000-8000khz
the crappy low rate audio will sound better, and compress better.
then maxamise the volume in the volume maxamizer, that will fix other poor audio issues.
all this can be Batched so it can occur enmass on a set of files.
this is NOT for MP3s (music only) as they sound great, the video playback needs more help.

Drop the video into Virtual Dub.
WHY
because virtual dub can tell a lot about a video,
it will tell you if there are issues of various sorts right from the start. with it you can edit by keyframe,
and save things out without de-re-compressing them ,
although in this situation just leaving the video completly decompressed is the task.

in AUDIO add back your WAVE file you extracted with gold wave,
do not use the source audio.


In Virtual Dub, add a mess of FILTERS, here is my list

Temporal smoother - this gets rid of moving noise between frames
and makes things cleaner for compression.

Levels - expand the range of the brightness and contrast of the video as widely as possible ,
beings its going to be a postage stamp the widest range of contrast and brightness is going to be nessisary.

HSV - with this filter i increase the saturation,
while dull greys might be usefull for a large screen DVD ,
were working with a few pixels and definition of color contrasts and general contrast is nessisary.

Sharpness - Again same thing applies, the watch has a Micro picture
you want to increase the contrast between elements sharpening will accomplish that.

Resize - resize to the exact size that your intending to output,
128x128 for a square pic, or 128x96 for your average 3-4 aspect stuff.
use the bicubic 1.00 style.

Crop - crop things early by adding cropping to the top filter,
this watch is SQUARE it is so opposite 20x11aspect wide screens it is rediculous,
you need to crop out everything you possibly can to make the dancing postage stamp.


Audio Filter
i use a Lowpass audio filter , depending on the video,
4000 cutoff with 64-255 taps or a 8000 Minimum for stuff with more highs and less noise

WHY
why do all that stuff? cause your trying to get a visable picture as small as possible none of it is Nessisary,
it just is what i use to attempt to get a fair pic on a SMALL item.


In ViDEO go to FRAME RATE, and check to see if the frame rate of the video matcheds the audio.
you may or may NOT want to have it adjust the frame rate (to the audio).
some videos have a dire need for a frame rate adjustment some work fine with no adjustement.
WHY
if a video is (what i call) MIGRATING out of sync,
in virtual dubb you can fix that by adjusting the frame rate in as small an incrament as .003.
this is because some idiot (like me even) forgot about the complications of "Drop Frame" video and other stuff (29.970).
If a videos audio is MIGRATING to an audio delay reduce the frame rate.
If the videos audio is migrating to Audio comming out before the video increase the frame rate.
(migrating means changing over time, it has nothing to do with always being out of sync)

Now DECIMATE (change the frame rate) the video frame rate HERE in virtual dub,
if you LET the smv converter do the decimation, it can cause audio sync problems again.
do not Decimate by x2 or x3 but SET A SPECIFIC frame rate that is exactally what you intend to be the final output.
Set your frame rate to (about) 1/2 the original file
the original file already has some kind of frame rate or field rate
they are usually (about) 24 25 30 or 50 and 60 interlaced fields
you want to reduce this an even ammount MINUS the fractional ammounts (.997)
so for 24s set it to 12 , for 30s and 60s set to 15 for 25 pick 12 or 13 or even 25, but 25 is hard for the watch to do.
WHY
Leave nothing to chance , or that stupid program,
because given the oppertunity the SMV converter will screw up.
by setting the exact size and frame rate here ,
virtual dub will do it without messing up the audio sync.
if you decimate the frames by x2 or x3 and the framing is 23.997 then your still going to end up with audio sync issues,
you want Vdub to do the decimation and you want it Exact numbers not fractions.

in AUDIO interleave set all the stuff in there , use 250 or 500 for the PRELOAD
and use a 1 Frame interleave. you can put in miliseconds, but there is no reason for it,
you want the audio attached to each frame, and that is what will occur even if you specify miliseconds.
WHY
interleave doesnt cost much in size ever,
it insures that the audio FOR that frame plays with that frame.

in AUDIO offset the Audio track by about 30-120ms or so,
depending on how badly the video was already out of sync,
and how slowly your video updates on your watch.
WHY
the video on this watch seems to be more tossed into the side ,
take a display and make believe you can do video with it.
hey its a MP3 watch and it does that,
everything else like the actual video seems like it was a passing thought of a geek connected to marketing :-)
the video is displayed Delayed from the audio, the LARGER the video is in KB/s the slower the video will be shown,
and the larger your audio offset delay should be.

Compression - none
go to the video compression and audio compression and just select none
WHY
your only trying to aqure a video that you can compress to SMV
there is no purpose in (again) compressing it to a format it will not be in.

Now , that was fun wasnt it :-)

EDIT the video taking out all extranious trash that will serve no real purpose,
commertials credits beginning titles, whatever.
WHY
because you cant edit IN the watch, you can only view a video,
you cannot scan past anything,
so this is your opertunity to remove what you would want to scan past.

Check check check: IN Vdub Check to see that your audio all matches ,
and that it doesnt migrate over time spot-check the audio sync and picture at the beginning the middle and the end.
you DO want a small delay in the video, with the audio playing before the video,
you dont want the quantity of delay to change over time.
What you see is mostly what you get in Vdub, unless there are bad video frames.

Save out as a SEGEMENTED avi.
in virtual dub when you go to save you can ask it to segement the video out into pieces,
use the number of frames to define a segement (because how long each segment is is important , not how large each is).
i use 8000 for 24fps videos , and 4000 for 12fps , and i output in and convert to 1/2 the original frame rates.
WHY
the watch has no scanning, if you output it in seperated hunks you can jump foreward or backwards to the next hunk,
if you missed something or got distracted.
if the video is only 3-5minutes to begin with :-) well then you dont need segements do you.
there is a short break (interruption) when the watch (automatically) moves to the next video segment,
you want about 5-10 minutes at most between segments.

when your done, you should have A video with these properties
10-15Fps (decimated)
128x128 or 128x96 pixels (resized)
uncompressed video, uncompressed audio
multiple seperate avis with 00, 01, 02, 03
the videos (each of them) should be in sync at the beginning the middle and the end , or (of course) slightly delayed.

ok, having fun yet :-) now its time to convert