I was testing and thinking some new things that can improve speed and quality when using TMPGEnc.
One thing was to predict CQ value much faster than the traditional way.
The method goes something like this:
1) Encode test clip in CQ rate control mode. Select save encode log, or choose veiw encoderlog right after you start the encoder
2) Sizefactor = WhantedFilesize / Filesize
3) Pred 1 GOP = Bitrate 1 GOP * Sizefactor * Avg Bitrate / Bitrate 1 GOP
§ "Bitrate 1 GOP" is the "Current bitrate" value you find in the Encoderlog after the first GOP
4) Next CQ = (Current CQ + (Current CQ*Sizefactor))/2
5) Then encode agian. This time you only need the first GOP, which is often only the 24 first frames.
-Next CQ = (Current CQ + (Current CQ * Pred 1 GOP / Bitrate 1 GOP)/2
And then keep on going intill "Bitrate 1 GOP" =(+/-5%) Pred 1 GOP
I haven't tried the method so very much (it's some work...), but it seems to work. If the method seems to work (but it may have some modifications???), it should be pece of cake to write a litle program wich do the boring job to calculate, read logs etc.
I'm though that it's possible to predict CQ in a fast way if you use the right calculations and use some values from the encoder log.
I also tried to encode the test clip with 2 pass vbr and use some values from it to calculate the CQ, but I haven't found any good way to do this.
An other exiting way to use the encoderlog, is to make avisynth script of them.
Ex:
You encode the movie with predicted CQ, with a very high max bitrate (ex 10 000 kbps)
When encoding is finished you go throug the encoder log, and GOP's with high bitrates, will used stronger avisynth filtering than GOPs with lower bitrates.
And maybe, in the future we use MVBR (Manual varible bitrate) instead of CQ. In this mode we can change Bitrate, Quality, Quantize Matrix, GOP setting, motion search type, etc during encoding. We can then squeeze out the best quality that is possible, and fill up a CDr to the edge!!!
SORRY MY BAD ENGLISH....