View Single Post
Old 17th October 2006, 22:55   #132  |  Link
DDogg
Retired, but still around
 
DDogg's Avatar
 
Join Date: Oct 2001
Location: Lone Star
Posts: 3,058
Quote:
Originally Posted by buzzqw View Post
i will analyze tomorrow rest of post

sorry ... but in multires message what width was suggested ?
704. Size had been set to automatic.
Quote:
you can read also in log file at line "AutoResize movie size:"
also: what final size have you got ?
Width or filesize? 704 was the width and I am redoing encode. later: The prediction in the log for video was 1,367,448,960 and actual was 1,307,182,261 ... 95.6% not bad at all
Quote:
autores will search WIDTH for an average CRF of 19.
??? I had manually set to 18 to MATCH the crf in the profile.
Quote:
Can you check this value with this tool http://www.64k.it/andres/data/x/x264...1.5-vr_nga.rar , it analyse the stats file
Will play with it. [What stats file? There is none with crf mode, is there? Btw, where did this tool come from? Are there any docs?]
Quote:
if in this message is suggested 704 then 704 is used... BUT i think i have understood your point of view...
Not sure which point of view you are speaking of.
Quote:
(idea for next version: add button auto REDO -> check for predicted size, if > SIZE requested, diminish width, redo again)
Maybe useful for some. I only use the autores to find a point it does NOT reduce width. In other words, if it reduces width I add a filter, change matrix, or add size. The very last thing I want is for the resolution to be reduced. It is kind of funny, using autores in a backwards manner allows me to find out how NOT to reduce resolution. Since I have input a crf value for the sample, this tells me I can get close to that crf value without wasting any space and without reducing resolution.

That is one of the reasons I think you should parse the crf value contained in a crf profile and put it into the sample box value. That crf is what the encode will be created with right? So for a crf mode encode it makes sense to start multires with a matching crf in the crf sample box. IF a user then changes the CRF value in the sample box (for a redo), does the redo, and then selects "no" to redo, the crf value in the sample box should be used for the final encode as that was the crf used to make the prediction. (only a crf profile). At least that is how I would do it and I think it makes sense.

Last edited by DDogg; 18th October 2006 at 03:51.
DDogg is offline