New MainConcept H.264 TechPreview (beta 2)

[LostMP4]

http://www.mainconcept.com/h264_encoder.shtml

New H.264 Preview #2 Features and Enhancements:

1. Baseline and Main profiles
2. Quarter-pel motion-estimation
3. The amount of reference frames is only limited according to chosen level/profile
(for example 5 reference frames for Main, D1-size)
4. Motion-search range is not limited by encoder (only by Profile/Level)
5. Configurable GOP-structure
6. Main Profile only:
6a. Interlaced (field stream) and Progressive (frame stream) Video
6b. Bidirectional-interpolation (B-slices)
6c. CABAC bitstream encoding

It doesn't work here :mad:
The program shuts down as soon as I click "convert" :confused:

[WaryWolf]

how strange, i was just about to play with the first version when i decided to check here first... :)

will post results later.

[JohnV]

Works here.

[Paced]

Originally posted by LostMP4
It doesn't work here :mad:
The program shuts down as soon as I click "convert" :confused:

Same thing happens to me.

[LostMP4]

By making multiple windows captures I saw it doesn't recognize the 3DNow capabilities of my Barton, then only says: starting the H264 encoder

[LostMP4]

I've a Athlon XP 2600+ (Barton, 11.5x333)
Nero Vision Express 2.1.2.18 installed
I've tryed to open a lot of different media files, codec and settings but the results it's the same

Grrr I can't post the screen!

MMX:y SSE:y SEE2:n 3DNowExt:n 3DNow:n CMOV:n
Initializating video encoder
Number of CPU: 1,available: 1
H264 Video encoder initialized (320x240)

Then it disappears

PS: what is CMOV?

[ivan_alias]

Doesn't work here, Win 2K

:(

[chilledoutuk]

dont work for me either

[ajp]

To those having problems, have you tried disabling or enabling "Directshow Import" under "Import Modules"? That helped sometimes with the normal Mainconcept Mpeg encoder. I see mine was disabled and it works for me... just a thought.

[Bulletproof]

I'm getting errors like these:

M089:Video DTS (7254.00ms) underflow at pack 1277 by 6.00ms
M089:Video DTS (8255.00ms) underflow at pack 1467 by 18.00ms
M089:Video DTS (8355.00ms) underflow at pack 1484 by 1.00ms
M089:Video DTS (12760.00ms) underflow at pack 2309 by 19.00ms
M089:Video DTS (15262.00ms) underflow at pack 2759 by 2.00ms

The resulting video doesn't even open in my Windows Media Player either..

[LostMP4]

Originally posted by ajp
To those having problems, have you tried disabling or enabling "Directshow Import" under "Import Modules"? That helped sometimes with the normal Mainconcept Mpeg encoder. I see mine was disabled and it works for me... just a thought.
Yes, I tried enabling 1 import module and disabling the others: AVI, DirectShow, MOV

I was thinking that maybe Nero's high merit filters could be involved.. it's faster to ask the configuration of a PC that can run this encoder

Any clues?

[Bulletproof]

Ok I got it to work but it only works for me if it is set to "H.264 Baseline", if I set it to "H.264 Main" I get a bunch of DTS errors and a file that doesn't decode.

[LostMP4]

Originally posted by Bulletproof
Ok I got it to work but it only works for me if it is set to "H.264 Baseline", if I set it to "H.264 Main" I get a bunch of DTS errors and a file that doesn't decode.

What kind of hardware did you used?
What kind of file did you convert?
Which codec are installed i your system?

[Bulletproof]

Ok here's whats happening right now, it seems short clips will decode fine, long clips will just not show any video. I converted a HuffYUV file. I got the Main profile to work also, but I had to reduce the bitrate to 1000 and the clip has to be short too. I am also using a P4. In any case, the results at 1000kbit/s is just a blurry mess.

[superdump]

I'm using an Athlon XP and it also doesn't detect 3DNow functions for me and it crashes shortly after clicking convert. Log file output...

H.264 video encoding
MPEG muxing
MPEG output file: C:\InProgress\clip1.mpg
Video source:

MMX:y SSE:y SSE2:n 3DNowExt:n 3DNow:n CMOV:n
Initializing video encoder
Number of CPU: 1, available: 1
H264 Video encoder initialized (704 x 288)

[lexor]

Considering that if you look around their site (especially SDK part for all of thier encoders) claims to Intel optimizations are plastered everywhere.
So far I'm seeing only AMD guys with problems, can Pentium 3/4 users comment, please?

CMOV is if I recall correctly an asm command (assembler code) though I very much doubt such code is the problem, they could've overtested on Intel platrom and ignored AMD.

[chilledoutuk]

Did they test this at all before releasing as too many people are having problems. my encoding machine is based on a amd xp if it is just a problem with amd's then that is pathetic software developmet.

in fact i have never been overimoressed with mainconcepts encoders which generaly underperform in more ways than one.

I am more interested in nero digital's h.264 encoder as they seem be on a roll at the moment when it comes to codec development.

[LostMP4]

Someone should try to contact them :(

[Longinus]

It works here...
Tried with a huffyuv source, and encoded only the video as Elementary Video, and played the MPV file just fine with Windows media player or MPC. I used main profile and 500 kbs, to try it with a low bitrate..

I liked what I saw... But I still prefer XviD with the same bitrate (at least for the video I tried).

[LostMP4]

Originally posted by Longinus
It works here...
Tried with a huffyuv source, and encoded only the video as Elementary Video, and played the MPV file just fine with Windows media player or MPC. I used main profile and 500 kbs, to try it with a low bitrate..

I liked what I saw... But I still prefer XviD with the same bitrate (at least for the video I tried).

AMD or Intel CPU?

[Longinus]

P4 with hyperthreding on...
You guys are probably right... I'll try with my AMD computer to see what happens.

[bond]

what about the decoder coming with it? is it still as buggy as the first preview, meaning showing a too big image size and a blurry picture in the beginning (but the later one is maybe caused by a fixed first frame quantizer?)

[Bulletproof]

Still blurry picture in beginning, and you can't turn off the loop filter. The implementation in FFDShow in my opinion is better and doesn't even have all those features they listed. There is supposedly an apple H.264 encoder in the latest Tiger MacOS beta, has anyone tried it? So far I'm unimpressed by all the H.264 tech demos.

[Andrey]

The same here:
When using Main profile numerous "underflow" (whatever it means) errors occured and file isn't playable.
The same result for baseline :(

Tried: clip 512x304, 6m34s w/o any sound...

[Disabled]

I can't use it either. I tried to use every import module (except DV, as it wasn't able to read my avi DV files) and disabled the others, because someone stated that might be an issue. I tried Main and Baseline profile, Pal/NTSC, different stream types, bitrate and much more and wasn`t able to encode a single frame.
Perhaps someone who got it to work might post his system specs.
I have and an Athlon XP, WinXP SP1 and no other H.264 codec on my system.

[Tommy Carrot]

Well, i have an athlon xp, and this encoder doesn't work here either. Apparently they couldn't even test it on an AMD machine, and this tells everything about the quality of their products...

[Bulletproof]

Andrey try only encoding 30 seconds or a minute of video only, the file seems to decode then. I guess it can be said for certain now that AMDs are not working, and even on P4s the program doesn't work 100% properly. My hope is with Nero and Apple to make a good H.264 implementation.

[Andrey]

>>Andrey try only encoding 30 seconds or a minute of video only
Ok, will try today's late evening and post results.

My machine is C2600, so it should :) work.
File I tried before was huffyuv, bitrate was set to 650Kbit.

[ivan_alias]

So basically this 'tech preview' is poo.

So many people having problems with it. H.264 is going to be a big deal and lots of companies are trying to get in their early and this is the result. Shame... Hope fully the bugs will be fixed soon as I really would like to try some H.264 in a full scale backup.

[Andrey]

The same for 30s huffyuv source, except no errors while encoding this time.
I've deleted this "demo". :devil:

[yodoso]

I have a athlon xp 2500, and main concepts h264 preview worked for me. I opened a video only avs file from an mpeg 2 source, It took 70 minutes to encode my 2 minute clip. Dark scenes didn't look too good, bright scenes looked better. My avs is attached. BTW it looks like their bitrate is calulated in KB/s instead of kbit/s because my file was way oversized. The result is an mpg file

[ivan_alias]

Has anyone got this encoder to work reliably? It seems such a shame that a decent company like this produced a test app that didn't run on many systems.

Anyone using it well?

Ivan

[LostMP4]

Originally posted by ivan_alias
Has anyone got this encoder to work reliably? It seems such a shame that a decent company like this produced a test app that didn't run on many systems.

Anyone using it well?

Ivan

They've updated their encoder page: only Pentium IV is supported

[bond]

for keeping things together: threads merged

[bond]

ok in my recent tour through the available avc encoders (yeah there is another world outside nero and x264 :D ) i tested the mainconcept codec with the following findings:

first of all maybe the most interesting thing:
the mainconcept encoder seems to be the first (and only) avc encoder to offer an advanced option of p-frame reordering (meaning storing the frames not in the "conservative" way, as done in mpeg-4 asp and other avc codecs. i will have to investigate this further)
this option can be regulated via the "reordering delay" option in MC

beaware the ffdshow and nero decoder seem to not handle reordered p-frame decoding correctly
mainconcept, moonlight and atemes decoder do

i couldnt test videosofts decoder as it only works with .avi and .avi (yeah the technology) cant handle p-frame reordering!

some specific findings about the encoder:

preview 1:
- works for all cpus (damn slow)
- max. 1 b-frame possible
- enabling b-frame sets the reordering delay to 1 (no matter what the gui shows), meaning no advanced p-frame reordering is used
- the encoder always uses 3 multiple reference frames, no matter what is set in the gui
- transport stream output is broken (mplayer tells me the file signals mpeg1), ES and PS output works fine
- the preview 1 decoder is buggy as hell (frames partly green aso...)
- ffdshow crashes when reordering delay is set to > 1 and combined with mref is > 1 (both seperated dont crash), nero decoder shows jumpy playback in that case

preview 2(a):
- only works with pentium4, why i couldnt encode myself (thx cruncher for doing a bunch of encodes for me! :) )
- gui says max 1 b-vop is possible, but defacto the b-frame including streams used 2 in a row
- enabling b-vops disables reordering still
- TS still broken (mplayer tells me the file signals mpeg2 this time)
- the decoder seems to have been fixed (also works on non pentium4)
- mref other than 3 are setable now
- ffdshow shows same behaviour as described above
- avc profiles and now even levels are choosable

[JohnV]

Originally posted by bond
beaware the ffdshow and nero decoder seem to not handle reordered p-frame decoding correctly mainconcept, moonlight and atemes decoder do No, Nero decoder does handle p-frame reordering properly. The problem with the jumpy playback with our decoder came from your MP4s which lacked the ctts table. But we did a workaround and now our decoder plays correctly also those kind of MP4s..

[bond]

Originally posted by JohnV
No, Nero decoder does handle p-frame reordering properly. The problem with the jumpy playback with our decoder came from your MP4s which lacked the ctts table. But we did a workaround and now our decoder plays correctly also those kind of MP4s.. interesting info, thanks! :)

[bond]

ok just for completeness mp4creator should support, starting with version 1.2.7, the correct muxing of such streams into .mp4 using such "not orthodox frame coding orders" with p-frames (the frames aren't coded in normal order ie you can code frame 3 before frame 4) by creating the till now missing ctts atom :)