Okay this something that stems form a common source of misconception tahts far more common with 3d users than comp artists, infact a acomp artist shoudl usndertand this.
Fisrtoff theres NO differnce between a clip and a image sequence form a comp apps point of view, they are just data, but because sequences usualy represent lower order data ist aoffcourse prefeered, if you happen to have it.
A movie file is jast a packaged form of a sequence, or in many cases it represents a compressed form where cpu is used to substitute for size. Now compressors come in 2 forms, lossy and nonlossy.
The term lossy refeers to the fact that human eyesarent all that accurate, so you dont necceserily see the difference even tough the image is mutated (easiets way to se this is to take a uncompressed image and sacve it out as jpeg and ajust gamma radicaly on that), the mutation is sualay caalled a compression artefact. Now the mutation stacks on previous mutations, so saving out files in compressed forms overa nd overa gain can eb a HUGE mistake.
Also for most comp operations he artefact can become a showstopper. Becaus ethe time you fight and fix artefacts can be more then the entire job would be in fisrtplace.
Nonlossy does not mutate but makes for MUCH (about 25-100 times) larger files.
Now for soem dowsides of codecs, remeber i told you they used more cpu. Now offcourse its a tradeoff for a larger file but tehnagain it takes more time to compute the frame out, now its hard to say where the optimum is because the optimal cpu/ disk lookup ratio depends on your disk read speed. Also most codecs (movie packaging/unpackaging routines), are optimized for sequential view, wich can mean that jumping around in the sequence can be much more costly than youd think, sometimes fetching the result farme can mean you need to fetch all 25 frames piror to the one your are at and then next 10 to know how the frame actualy looks like. Incidentaly all this eats memory too.
Aslo usualyu compression does not fare wery well in pararlell processing of the result, wich sequences usualy do, so 10 computers could do teh sme job 10 tiems faster in a sequence dump.
Now finding the optimal worklow in this minefield of interlinked issues is a bit hard to understand. And theres no straight forward answer.
just 2 rules of thimbs.
1) use unlossy formats as loong as you can prefeerably the entire way untill the final output, also store the uncompressed footage for archiving.
2) compression takes time, tighter pack usualy takes more so inorder to save cpu cyles in the pipe try to compress as littel as you can on the way there. Then compress all in one go realy well (and many times to find the optimum) Se theres no point in wasting time to copress something thats going to be uncompressed a moment later.
