Avisynth 2.03 released!

[sh0dan]

Just uploaded Avisynth 2.03.

This release has many new features, optimizations, improvements and bug-fixes.



The final changelist:
-------------------------
- Fixed EMMS bug in ColorKeyMask
- Fixed YUY2 detection issues in Layer
- Added friendly error message for Layer ops
- GetLeftChannel - Returns left channel.
- GetRightChannel - Returns right channel.
- MonoToStereo(clip1_left,clip2_right) - muxes two clips together as one stereo clip.
- MixAudio(clip_1,clip_2,float clip1_volume, float clip2_volume) Mixes two audio sources together.
- Added ConvertBackToYUY2() That only uses chroma from the left pixel to avoid shifting chroma color by multiple YUY2->RGB>YUY2 conversions. ConvertToYUY2 remains unchanged.
- Added Documentation for new parameters
- Added sound to Loop() function.
- New function ConvertToMono(clip) - convert stereo to mono.
- New function KillAudio(clip) removes audio from clip. Use this if you get crashes with compressed AVI sound.
- new filter: ColorKeyMask + bugfix in Layer
- New function Normalize(clip, float max_left, float max_right). Normalizes audio. Both floats are optional. If maximum values (0 to 1) are used, the stream will peak at this level, otherwise the peak will at full volume.
- TemporalSoften (MMX & C ) (previously released as plugin TemporalSoften2)
- MMX optimized HorizontalReduceBy2() in YUY2 mode. MMX is now pixelperfect in compare with C implementations.
- Changed order in ReduceBy2() since VerticalReduceBy2 is faster.
- VerticalReduceBy2() has been MMX optimized. More than twice as fast. Filter checks if image is too small to be reduced.
- Colorbars() now also generates a test tone. Test tone is a 440Hz sine at 48KHz, 16 bit, stereo. The tone pulses in the RIGHT speaker, being turned on and off once every second.
- ResampleAudio() now skips conversion, if samplerate is already at the given rate.
- No athlon codes in merge.
- Converttoyuy2 MMX optimized.
- Fixes in greyscale() in RGB mode.
- Optimized layer() and decomb-filters.
- Compare(clip filtered, clip original, string channels="", string logfile="", bool show_graph=true)
- AssumeTFF() and AssumeBFF() to force parity
- Documentation restructured and updated.
------------------------------------


Get it at Sourceforge (http://sourceforge.net/projects/avisynth2/)!

Now we should get our minds on the 2.1.0 release that will hopefully have compressed audio support, float point samples, a nice website and a lot of other stuff.

[Dreassica]

Great, let's try it out right away!!:cool: :cool:

[Marc FD]

BTW, when would we see optimized BitBlt ??
or it's a definitive bad idea to optimize it ?
(i read the thread but did not see any global "conclusion")

[Richard Berg]

Workin' on it -- I'll have another benchmark app for people with various CPUs to try in a couple days, then I'll decide what to do about it.

[Marc FD]

great :)
if you could get it much much faster i think we should see a good speed improvement
something between +0,01 fps and +100 fps :)

(yes i don't know the impact but i want to see it working)

[sh0dan]

Benchmarks so far are very theoretical, since BitBlt is used in so many filters in many different ways, so it highly depends on your filter chain.

[Didée]

Er, sorry ...

I´ve read about that "BitBlt" thingy quite some times now.
Could you give a short hint, for the dumb of us that are no programmers at all (but nevertheless have some natural technical understanding), what that´s about? So that we can imagine - something ...

[sh0dan]

BitBlt is used for copying larger amount of data, in graphic terms, blitting. It is used to create copies of images, and other tasks, where filters needs to move larger amount of data at the same time. Now, Avisynth is using default C-routines for this task, but Richard is looking at some MMX/ISSE optimizations that provide better cache/memory bandwidth usage. He's also looking at a routine, that fills a piece of memory with a given byte value.

[sh0dan]

2.04 is now released. It contains a bugfix, and compressed audio support!

[Marc FD]

let's test :)

[cult]

whenever i use the version()
it shows me v2.01 4june 2002

[Richard Berg]

I'd rag on Klaus again, but I did the same thing in v2.02 ;) It's fixed in 2.04.

[sh0dan]

Can't be! I corrected the source, and it's even in the binary file (Search for "Avisynth v2.0.4"). You must be referencing an old avisynth.dll! Phew you just got my heart going there! :scared:

[MaTTeR]

Originally posted by cult
whenever i use the version()
it shows me v2.01 4june 2002

Happened to me also until I finally discovered a 3rd dll hiding in the Vdub folder:D

[cult]

works now.Shows correct version 2.0.4 07 august 2002

[LigH]

Just a question about the relation between two teams: How close are the developers of AviSynth2 (directly at SourceForge) and those of the "original" line at VideoTools? Will both lines drift away in the future (like DVD2AVI 1.77.x and 1.8x), or are sources exchanged? I just wonder, if the BicubicResize chroma clipping becomes fixed in AviSynth2, will it also become fixed in the VideoTools release? I reported both teams about that issue - the AviSynth2 team already confirmed, the one at VideoTools not even answered until today...

[sh0dan]

@Ligh:

(This is a sensitive political issue, so there may be more views on this - Richard and dividee has the more accurate story)

Avisynth2 was created because the original Avisynth project on Sourceforge is dead, and development has been stopped. So avisynth2 is the only open project, where Avisynth is still being updated by the community. Videotools.net does not seem to be updating avisynth - all new features (link2, compressed audio, optimized conversion) has been pay-ware. Videotools.net could update their avisynth with the new features, but they haven't done so far.

[Richard Berg]

That's precisely the point; the Videotools branch is dead, in my opinion.

When I emailed Edwin "videotools" Eggelen about 6 weeks ago, he confirmed his plans were to perhaps make another patch or two to Avisynth but generally to focus on selling plugins. No doubt he's aware of what we're up to by now, but I haven't heard from him since 7/1, a couple days before I started the new SF project page. The new project & 2.xx numbering were at first mostly for technical reasons & to avoid confusion, but it's clear enough now that leaving Videotools behind was for the best in the long run.

AFAIK other than Edwin there is no 1.xx "team" -- Ben left ages ago, and everyone else has been added to 2.xx if you look on SF. Haven't heard from some of the old avisynth-dev-list crowd in awhile (hero, steady, Mathias), but Ole Hansen got ahold of me recently to saying he wanted to commit an enhanced ConvertFPS to Avisynth2 soon; meanwhile, all the Doom9 regulars certainly seem to be on board.

[LigH]

Okay - thank you for pointing those facts out. Now I wish all the AviSynth2 team members good luck in hunting this issue down! (I was at first complaining Tweak to be buggy and had to apologise Donald Graft after finding out that instead BicubicResize was the reason...)

[int 21h]

Just want to say, great work. I'm really excited to see Avisynth in open development again away from the payware bull*@#@. You guys are doing a great job so far, keep up the good work.

(Maybe after you get bug hunting and optimization stuff done, you could investigate just how Link2 makes a compatible Avisynth file for CCE 2.6x and implement something similar ;) )

Give 'em hell, best regards.

int 21h

[-h]

(Maybe after you get bug hunting and optimization stuff done, you could investigate just how Link2 makes a compatible Avisynth file for CCE 2.6x and implement something similar ;) )

If it really had to work, one could write a "special" version of avisynth that created a hook to the file system API and intercepted any call CCE made to the file in question, filling the read buffer with the appropriate data. Then it really "would" be an AVI file it was reading. No way it could turn that down.

There's probably a neater way, but I have no idea what that would be.

-h

[Blight]

Considering that CCE 2.66 is supposed to be able to natively read AVS files, I don't think this should be an issue much longer.

[sh0dan]

Originally posted by -h

If it really had to work, one could write a "special" version of avisynth that created a hook to the file system API and intercepted any call CCE made to the file in question, filling the read buffer with the appropriate data. Then it really "would" be an AVI file it was reading. No way it could turn that down.

There's probably a neater way, but I have no idea what that would be.

-h

If you were able to do so, I think it would be a great, great feature. I know cce will have the avs-thing fixed, but it would be a very good thing to be able to use avs for other stuff than just encoding. I think it would be a very good thing to be able to use it for premiere, after effects, avid, cleaner and other software, I use at different times. A real AVI solution, that isn't dependent on the client solution would be a great leap forward (like compressed sound is).

Would you be able to find time to implement it on top of the current CVS source, and release a test version?

[-h]

If you were able to do so, I think it would be a great, great feature. I know cce will have the avs-thing fixed, but it would be a very good thing to be able to use avs for other stuff than just encoding. I think it would be a very good thing to be able to use it for premiere, after effects, avid, cleaner and other software, I use at different times. A real AVI solution, that isn't dependent on the client solution would be a great leap forward (like compressed sound is).

I could try it, but not for a week or two yet. The container code wouldn't be too hard (i.e. the file system hook), but I'm not too familiar with the content of an AVI file to write the rest of it. Perhaps someone else could fill that in.

-h

[-h]

If you were able to do so, I think it would be a great, great feature. I know cce will have the avs-thing fixed, but it would be a very good thing to be able to use avs for other stuff than just encoding. I think it would be a very good thing to be able to use it for premiere, after effects, avid, cleaner and other software, I use at different times. A real AVI solution, that isn't dependent on the client solution would be a great leap forward (like compressed sound is).

I could try it, but not for a week or two yet. The container code wouldn't be too hard (i.e. the file system hook), but I'm not too familiar with the content of an AVI file to write the rest of it. Perhaps someone else could fill that in.

-h

[Koepi]

heh, a glitch in the matrix! what did they change? 2 posts with the same content ;)

SCNR,

Koepi :)