Video Enhancer next release

[Marcel-78FR]

Hello, when the next version with CUDA capabilities will be released?

Marcel

[Dee Mon]

Short and honest answer: we don't know yet.

[Manoman455]

I thought video enhancer 1.9.4 was supposed to come out months ago, whats taking, i would really like to get my hands on the trial and give it a test run considering its supposed to be faster thanks to new colorspace. also please fix the issue where video enhancer crashes PC if not running in compatibility with desktop composition disabled.

[Dee Mon]

Sorry, the delays were caused by battling with multithreading problem described in the blog:
http://www.infognition.com/blog/
and later by a couple of business-to-business contracts which took our time.

[Manoman455]

So theres not release date?

[themaster1]

We are August 26th ,where is the New version ??!!!!

[Dee Mon]

Number of changes got a bit higher than we expected, but now almost everything is ready, VE 1.9.4 is in the phase of final polishing and testing now. Couple of screenshots:





Changes in 1.9.4:
* YV12 mode for super resolution - approx. twice faster than RGB32. Turns on automatically when the source video is decoded in YV12 unless user explicitly chooses to work in RGB32.
* Scene change detection (less artifacts).
* Higher overall SR quality and speed, thanks to new in-frame upsizing method.
* More precise control on SR quality/speed trade-off.
* Unicode, multilanguage support,  easy to add translations (just add a text file).
* Can work with lower than normal priority - doesn't slow down other applications even with 100% CPU load.
* Better Win7 compatibility.
* Some minor changes (like drag'n'dropping files to main window) and bugfixes (e.g. placing output files in same folder in batch mode).

No changes concerning input/output formats yet.

Release date: September 1st 2nd  .

[Manoman455]

Can we get quality comparison shots of VE 1.9.3 (high quality/RGB) and VE 1.9.4 (slide bar to far left/YV12)?
and is there a price point yet, or do we wait till release?
Thank you.

[Dee Mon]

Here's a sample of 2x upsizing of a file which was originally in YV12 (640x360 -> 1280x720, 300 frames).

Version 1.9.3 in high quality RGB mode worked 46 seconds:
http://stuff.infognition.com/mnt-ve193.png

Version 1.9.4 in YV12 fastest mode - 13 seconds:
http://stuff.infognition.com/mnt-ve194a.png

Version 1.9.4 in YV12 high quality mode - 25 seconds:
http://stuff.infognition.com/mnt-ve194b.png

Version 1.9.4 in RGB high quality mode - 37 seconds:
http://stuff.infognition.com/mnt-ve194c.png

Whenever any of them processes in RGB32 mode an YV12 file the video gets converted from YV12 to RGB32 by some standard system module (either Color Space Converter or AVI Decompressor, need to check) and a checkerboard pattern appears. This doesn't happen in 1.9.4 when it works in YV12.

All registered users will get the update for free. For new users the price will remain the same.

[Dee Mon]

Version 1.9.4 is finally available for download.
In addition to changes listed above now one can save and load projects - sequences of SR and VirtualDub filters with all their settings. Just right click the filters sequence list.

The manual is still for previous version, it will be updated a bit later.

[themaster1]

Ok i'm testing this new version 1.9.4 and i have spotted one BIg annoying thing, take a look:



Uploaded with ImageShack.us

I'm encoding a flv decoded with Fddshow in YV12 (no filter used in FFdshow)
As you can see in the result preview window there is some kind of "grid" for lack of better word it's very thin but you can see it clearly if you zoom-in.This grid doesn't appear in the original content so something's wrong !

The original video is 400x304 (codec On2VP6) i'm resizing  it one time X2 (best quality) and then finally 960x720 (speed mode) (so two SR filters) and just before the SR filters i use MSU smart deblocking 0.8 (latest version)

What's your opinion about that ?


Zoom-in:


Uploaded with ImageShack.us

[themaster1]

Well i think i found the culprit it's the filter Msu Smart deblocking, i suppose it's only compatible with RGB. I use YV12 in Video Enhancer.

[Dee Mon]

That's exactly what I described earlier:

Whenever any of them processes in RGB32 mode an YV12 file the video gets converted from YV12 to RGB32 by some standard system module (either Color Space Converter or AVI Decompressor, need to check) and a checkerboard pattern appears. This doesn't happen in 1.9.4 when it works in YV12.

While everything is in one colorspace, all's fine. But if you start mixing them conversion brings this pattern. We'll try to fix this soon by using our own conversion instead of relying on the system one.

[Dee Mon]

Updated version is available with new help and some bugs fixed, everyone's advised to reinstall.

[Sunny666]

I'm trying new version on the dual xeon server and have noticed that it doesn't utilize all computing power:



Only about 50% CPUs are used in general. Is Video Enhancer optimized for no more than 4 threads? There were checked "Work in RGB" and "Quality" options. Input file was avs script of DVD9 video with yadif as deinterlacer and color correction.