HCenc 023 released

[Guest]

Hank never gave me a solution for this and without the source code I am powerless.

[QuadraQ]

I should have stated that this only seems to be a problem with the NV version of DGAVCDec. When I tried it with the "normal" DGAVCDec it works perfectly fine.

[hank315]

It will be solved in the final 0.24 release, I can replicate the issue using DGAVCDecNV.
It's a complete lack of time ATM to get a new release out

[G_M_C]

Finnaly found the time to do a BD -> DVD conversion, using HCEnc & DVD-lab. The encoder works perfectly, as usual; But is was kind of funny.

It's been some time that i've done such a project, and i've upgraded to a Core 2 Quad Xtreme 9650, wich runs at a conservative 3200 MHz (400 x 8). The funny thing is that the whole 2-pass encoding process of a 2 hrs movie took less than the 2 hours the movie is long. It ran at more than 150 fps during encode

I remember times that encoding went @ 5 or 6 fps. That goes to show how technoligy has evolved in the years, but also that HCEnc seems to have matured considerably during that time as well. And that not oly gos for the speed of the encoder, but the quality it produces as well; Thats very very good (lum_gain=1 and AQ=1 seems to be a very good combo). So thanx again for all the work and time you have invested in the years !

[halsboss]

Originally Posted by QuadraQ

create a dga from a film avc mkv file, and then create an avs file ... Open it up in HCgui 0.23 fine, but when I press encode it's stuck. Tried opening up the HCenc 0.23 directly as suggested and it runs through the first pass fine, but then freezes on the second pass.

Oh NO !

Originally Posted by hank315

It will be solved in the final 0.24 release, I can replicate the issue using DGAVCDecNV.
It's a complete lack of time ATM to get a new release out

Oh Good. Thank goodness, this is significant, DGAVCDecNV is of no use to me otherwise. Is it still a problem when using the lossless intermediate file ?

Any timeframe on it as yet ?

[Guest]

You can use version 0.22 as I do.

Unfortunately it's an HCEnc issue. I tried to work around it but there is no way.

If Hank wants to lend me the source code I can try to fix it, otherwise we are totally at his mercy.

Don

[halsboss]

Thanks Neuron2. OK.

It it works for a 1st pass for QuadraQ, I wonder if it'll work when an intermediate lossless file is used for the 2nd pass to "work off". Will clutch at that straw tomorrow

Hank, you into sharing, matey ?

[hank315]

Here's a new HCenc 0.24 beta:

- added *INTRAVLC command
- added 4:2:2 input / output
- fixed some issues using AVCsource input

This will fix the problems using DGAVCDecNV.

[Boulder]

Thanks a lot, hank It feels weird that v023 was released a year ago..time flies so fast!

[Sharc]

Originally Posted by hank315

- fixed some issues using AVCsource input
This will fix the problems using DGAVCDecNV.

Thanks, hank!

[halsboss]

Hank, you are a Legend, mate.

[manolito]

Thanks a lot, Hank!

I can confirm that the -1pass and -2pass command line parameters are now working...

Cheers
manolito

[Fr4nz]

Thanks Hank for your efforts!

Excuse me for this question: did you see this thread? Are you planning something in the upcoming 0.24, in order to improve encoding in "dark areas"?

[kumi]

Thanks for the new beta!

Can hank315 or someone else explain in layman's terms what INTRAVLC is, and what it's good for?

[hank315]

About IntraVLC...

MacroBlocks can be encoded in two different ways:

-------------------------------------
| Intra VLC	    |   0   |   1   |
|------------------------------------
| Intra blocks      |  B14  |  B15  |
|------------------------------------
| Non-intra blocks  |  B14  |  B14  |
-------------------------------------

B14 and B15 are two tables which are used to perform the VLC: Variable Length Coding.
The outcome of this will be written in the video bitstream.
Table B14 is more efficient when there are only a few values in the 8x8 block, table B15 is more efficient when there are lot of values in the 8x8 block.
So in general table B14 should be used for low bitrates, B15 for higher bitrates.
IntraVLC can be changed for every frame.
So *INTRAVLC 0 means: always use table B14, *INTRAVLC 1 means always use table B15, *INTRAVLC 2 will use the VLC method which gives the lowest nr of bits for that frame, *INTRAVLC 2 simply does both and picks the table which generates the lowest nr of bits to represent the encoded block.

[scharfis_brain]

How much gain in bitrate do you estimate by Automatic IntraVLC compared to fixed IntraVLC?

[hank315]

The table shows file sizes of a testclip for different INTRAVLC settings using different Quantizers.
All other settings are the same so it shows the efficiency of the INTRAVLC setting only.

---------------------------------------------
| CQ | intravlc 0 | intravlc 1 | intravlc 2 |
|----|------------|------------|------------|
|  2 |   101674   |   100606   |   100576   |
|  3 |    53550   |    53140   |    53086   |
|  4 |    45531   |    45247   |    45182   |
|  5 |    33787   |    33782   |    33629   |
|  6 |    28421   |    28541   |    28330   |
|  8 |    23530   |    23844   |    23507   |
| 10 |    16561   |    17036   |    16561   |
---------------------------------------------

It shows table 14 (intravlc 0) is better at low bitrates (high Q) and table 15 (intravlc 1) is better at high bitrates (low Q).
Intravlc 2 (auto mode) always produces the smallest file size and converges at low bitrates to table 14.

So the gain using auto mode is pretty small... but it comes at almost no extra cost, about 1% extra encoding time.

[doxville]

Hi hank315, many thanks for all the time you spend on this great encoder!

Since 0.24 is a beta, I wanted to ask you, if you had the time to look over this issue, because I found the same behaviour as in 0.23. Would be great if that could be fixed in the final release (i am still sticking to 4:3 encodings), even if it's not a big thing to change AR yourself.

[cweb]

Just to thank you Hank for your new beta release..

[zack_ind]

Originally Posted by hank315

Here's a new HCenc 0.24 beta:

- added *INTRAVLC command
- added 4:2:2 input / output
- fixed some issues using AVCsource input

This will fix the problems using DGAVCDecNV.

Adaptive Quantization even when set to 0 doesnt work. In the ini file its not there but once we encode the log file shows it as 2.
This is in 0.24 beta , 1 pass vbr is quite brilliant