Real media to another codec

[MetalheadGautham]

Real Stuff

I want to get rid of my several real media videos, to make them more playable. What format should I convert them to? They are all usually 32 kbps audio, or ~170kbps video. some of the videos have cook audio. As cook is a part of ffmpeg, I am planning to preserve the cook audios.

What format of video is ideal to convert RealVideo to?

I am going to use a matroska container so there will be no problem of audio/video codec mismatch.

[MetalheadGautham]

Q. How will I transition my old VP3 files to the new format?
Because Theora is a 'superset' of VP3, tools can easily be created that will allow VP3 files to be losslessly transcoded into Ogg Theora format with no loss in quality.

thats how I want my real video to be converted - NO loss in quality, no gain in size.:readfaq:(lols)

[joseph5]

thats how I want my real video to be converted - NO loss in quality, no gain in size.:readfaq:(lols)
It can't be done. Every conversion has quality loss and/or bigger filesizes.

[mgh]

mkv takes real video and audio streams.demux your real media videos and mux to mkv. look in the real video sticky for utility to demux real files.

[joseph5]

I believe there's no need to demux first.

[pandy]

It can't be done. Every conversion has quality loss and/or bigger filesizes.

in theory it is possible that we know everything about source and codec and we can recode/transcode without loose any information and if both methods are comparable also without having bigger size. This process sometimes is called 'source reconstruction'.

[foxyshadis]

Source reconstruction falls under the same category as blind deconvolution and cryptographic hash reversal; even if the transform algorithm is perfectly known (which in the case of real it isn't, iirc, and in the case of deconvolution is almost always wrong), an arbitrarily large number of possible inputs can create the one output, each of which can only be found by brute force requiring an unfeasible number of cpu cycles, and randomly or having a human choosing between them. Hash reversal has the advantage that the input will follow some kind of syntax, usually, whereas trillions of nearly-identical videos may fit, differing only in the amount of grain/detail and placement of edges (or for low quality video, potentially differing wildly in detail structure). It'd be simpler and much faster to reconstruct a movie in photoshop than to regenerate it by brute force and human decisions.