File size prediction sometimes not accurate using vmesquita?
 

I'm currently using vmesquita's quide and calculator to predict file sizes but sometimes the total of the final sizes turned up to be much smaller than expected. Take for example, I was trying to encode a KDVD of 325 minutes video with 160kbps audio. Using vmesquita's calculator, it gave a sampler size of 40.79 MB. I tweaked the Q factor in CCE until the sampler size is 40.5MB which is pretty close and I always do it smaller since I don't want to face the situation where it can't fit onto the DVD-R. But now the actual final size of the movies along with audio add up to 3.6GB which leaves a lot of space to fill in. Is there a more accurate way to predict final video file sizes? I'm converting some episodes Divx cartoons, the only difference with vmesquita's script is that I deleted all the filtering lines.

I guess your problem is that your movie is too long.
I don't know the Vmesquita'a method but generally we take a constant number of frames within all the movie length to make the encoding sample. If you take for instance 2000 frames in a file a 120 min, you will be more accurate than if you take 2000 from a file of 360 min.

OK

1) Total length of source in seconds

2) Length of the sample in seconds

3) Divisor (divide step 1 by step 2 to get this number)

4) Your video budget (subtract the file size of your audio file from 800)

5) Target sample size (divide step 4 by step 3)

6) Hit that number, and encode.

------------------------------------------------------------------------------

My last two encodes have both landed at EXACTLY 800mb using this method. That isn't unusual for me. Until someone comes up with a tool that can duplicate my results, I'll stick to this...

That's exactly what I wanted to point out !
As generally the sample length is always the same (for instance, 2 min with CQMatic), the more your movie length is, the less you are accurate.

Are you sure you are doing the GOP factor right? Please remember that after determining the GOP, you must use it in the calculador and in the sampler line! :D I never got a error bigger than 3%, so I think this result of yours is very wired... :(

[]'s
Vmesquita