What's The Best Dial-Up Video Codec?

[augman000]

What is the best codec to use for Dial-Up streaming purposes? Which codec produces best quality and compatability and with what settings (bitrate, size, fps, etc).

[Sirber]

Real is pretty good for streaming.

http://www.realnetworks.com/audience/index.html

[buzzqw]

i think the Search(tm) Codec :D

BHH

[Manao]

I would have said the #12 rules codec.

[zambelli]

Are you planning on streaming it or putting it up for download? If you're streaming it, you're looking at max 40-50kbps and your best options are WMV and Real Video. If you're not streaming it then any codec optimized for low bitrates will do.

[IgorC]

violating rule 12 of forum.

However
1.Winamp5.1.x VP6.4.xx + HE-AAC2 . VP6 is better than RV10 (but RV10 better for anime), WMV9. HE-AAC2 is better than rv10 he-aac1 and wma 9.1.

2. Winamp has a midle size of RealOnePlayer10 and WMV9.

If QT7 will support HE-AAC1-2 it will be excelent player for stream.

[bond]

there are lots of codecs for that purpose and i am sure noone here made a comparison an a valid number of samples to be able to say what is the best codec
also whats best is highly subjective

so i would say use the ogg codecs (that is theora for video, speex if you use human voice, vorbis if you use music)
you can play such streams in windows media player (and any other dshow player) and realplayer with plugins (and support in qt is worked on)

[DeathTheSheep]

I'd suggest On2 VP7 or Flix (flash based). VP7 is the highest-ranking public codec in the world right now (as far as I know) in terms of SSIM quality measure. It's also new, and it's highly optimized for extremely low bitrate streaming content. It's also very easy to use. A personal favorite of mine, new and powerful VP7 has yet to be beat (or even matched), when it comes to streaming video.

At least try out the VP7 personal edition (free) and see for yourself! Licensing costs a bit of dough, but the quality easily does it justice.

[Manao]

DeathTheSheep : http://forum.doom9.org/showthread.php?t=90784&highlight=ssim+codec

VP7 is ranked lower than x264 for SSIMIt's also newThat's not really an asset, is it ?

[DeathTheSheep]

Woah, surpassed already! Wow, the progress x264 is making is truly astounding! The question is: can it be streamed as easily?
Sorry about the misinfo... it was tops until recently.

Higest performance codecs: VP7 vs NDAVC vs x264
VP7, Ateme AVC or x264 are really better than the others codecs. VP7 seem better than Ateme H264 Main Profil for very high quant (low quality or "low bitrate") and Ateme H264 Main Profil seem better than VP7 for lower quant (high quality or "high bitrate"). x264 High Profil is the best codec actually available. Ateme H264 is by far the best codec in this test ...
Yeah, streaming would be great with VP7. It truly does look a heck of a lot better than AVC (to my eyes) at streaming rates (25-60kbps).

Still, it's no longer top dog for DVD-rip-style encodes, at least, according to the metrics. *shrug*
Yeah, x264 is even newer (as new as possible...updated tons on a daily basis, still not officially released).

[guada 2]

Hey, Manao & DeathTheSheep

Maybe, it will be necessary to put you okay. :)
Which codec preserve maximum quality VP7 or x264?

[DeathTheSheep]

x264 typically preserves more detail and metric quality at higher bitrates (non-streaming), and VP7 preserves more with streaming.

SO WE BOTH WIN, HAHA! HA! Ha! ha. ......ha?
...A little bit too much, eh...?

[IgorC]

Last revision 375 of x264 is better than vp7.0.8 for all bitrates. There was a grat progress since then ... --b-rdo +0.5 SSIM , trellis 2 +0.2 psnr , subme 7 and more other new features.
and Ateme is best videocodec right now.
Also vp7 eats more cpu than h.264 . For x264 it's possible to use light settings for lower CPU usage .. no bframes, lower inloop filter, no cabac but to try compensate that with slow RD features.

vp6 already requieres quite fast CPU. Not all users have P4 for vp7 stream.

[bond]

VP7 is the highest-ranking public codec in the world right now (as far as I know) in terms of SSIM quality measure.this sentence sounds borked :rolleyes:

"its the worldwide one and only best of all times in human history ever existing codec (afaik)" :D

[DeathTheSheep]

"its the worldwide one and only best of all times in human history ever existing codec (afaik)" :D
WOW!! That one's exactly right! :D

LOL, maybe I overdid it a tiny bit? Fine, it's the "second" best or "third" best or something...;) Ya know, it has to be something like that... :)

[IgorC]

Bond , It will be great if there will be a new rule about troll like on hydrogenaudio forum. People gives their opinions without any relation with real situation.
If somebody wants to say what´s better he/she must provide some information - test, metric results anything. H.264 is better than vp7 and there was a lot of testing around it.

[IgorC]

Just because of curiosity
vp7.0.8 and x264 rev375 at 49 kbit/s http://www.mytempdir.com/276881

[Mug Funky]

@ igorC: it would be much harder to enforce that particular HA rule.

ABX is easy to do with audio, but for video it's particularly difficult (until some kind soul comes up with an abc-hr style app that plays any video).

also, video compression tends to exhibit more artefacts than lossy audio (WMA excepted... it's geniunely bad at all bitrates. confirmed with ABX :)), so it can make it easy to spot the "character" of artefacts and correctly identify the codec from that.

[bond]

@ igorC: it would be much harder to enforce that particular HA rule.

ABX is easy to do with audio, but for video it's particularly difficult (until some kind soul comes up with an abc-hr style app that plays any video).

also, video compression tends to exhibit more artefacts than lossy audio (WMA excepted... it's geniunely bad at all bitrates. confirmed with ABX :)), so it can make it easy to spot the "character" of artefacts and correctly identify the codec from that.exactly, thats why our rule is "there is no best"

[DeathTheSheep]

I dunno 'bout you guyz, but I'm havin a heck of a time tryin' to stream anything at over 32kbps...

So I don't think 49 is an option here; most dialup modems can't hold 50kbps for over 5 seconds anyway! In fact, in my particular case, my 56k maxes out at 45.2kbps max peak rate--it can't actually reach 56.6kbps, and neither can anyone elses. I think the industry max is 54kbps, and even then, good luck trying to stream something at 50kbps! :D

[DeathTheSheep]

Also vp7 eats more cpu than h.264
Actually, it seemed about the same to me as of 7.0.8 (drastic speed increase since 7.0.4 and below, when speed was measured). It's about the same decoding complexity, and I can decode it on my Pentium 2 like x264 (both for lower resolutions, of course). And if you turn off postprocessing, you can save a lot of processor time. Besides, the PP level adjusts itself according to your CPU load. A quad xeon rig would use different PP than my Pentium 2.

As for those 2 samples in IgorC's post, I thought they looked about the same. I didn't like the odd B-frame mush of x264, and I didn't like the smearing on the VP7, and the bitrate was still too high for dialup streaming--even without sound. After tuning the PP filter a bit, the VP7 one looked nice, and after substituting the inloop with SPP, x264 looked nice, too.
I'm curious, though, how the *real* streaming bitrates would hold up. Maybe turn off B-frames, use a frame decimation factor of 1, and try it again at 32kbps.

Or maybe....(light bulb turns on)....I could do it myself....(light bulb goes off)....but I can't download the trailer....
Cheers!

[IgorC]

Actually I was talking about vp7.0.8. That´s still eats more CPU than x264.
ok. 49 kbit/s was very high bitrate for you. So provide a test at 32 kbps and don´t forget about original. I don´t see any test. I don´t beleive your words and I have not to.

[DeathTheSheep]

Don't worry, man! Chill, ok? ;) I got my opinion from encoding and watching them for myself. Everyone's can be different, right?
I don´t beleive your words and I have not to.
Don't then. *shrugs* But I already know that, so telling me means nothing--I'm just stating my opinion, and everyone is entitled to thier own.

Yeah, x264 plays at 640x480 55% of my Celeron D 2.93ghz CPU
and VP7.0.8 plays at 45%. What, do you want a screenshot of my graph? I just made a test.
http://www.myfilehut.com/userfiles/5120/graphs.PNG
VP7 averages to 45-48% CPU, x264 averages around 60% CPU. C'mon man, I'm not pulling this out of my butt or something :p
And I also said that even if it was surpassed by x264 in SSIM, the 2 files you had looked the same to me with optimal decoding.

[Manao]

DTS : ah, but there, since there's only one decoder for vp7, but more than 10 for h264, you should mention which decoder / player you used :)

[DeathTheSheep]

Oh yeah, ooops... I used ffdshow with standard, clean x264 decoding (no added PP or filters), and...eh...vp7 for vp7 decoding ;)

PS: Optimal decoding=x264 with SPP level 5 deblocking, but I didn't use that in my test--just standard, pure x264 with inloop. For my CPU test, VP7 used PP level 3, and no CPU throttle, etc. Same background processes (only IE and virtualdub, i believe). All in all, I'd say this is a fairly accurate test.

[bond]

Yeah, x264 plays at 640x480 55% of my Celeron D 2.93ghz CPU
and VP7.0.8 plays at 45%. What, do you want a screenshot of my graph? I just made a test.
http://www.myfilehut.com/userfiles/5120/graphs.PNG
VP7 averages to 45-48% CPU, x264 averages around 60% CPU. this comparison is totally useless if you dont tell us what decoder you used. cause obviously better decoders decode the same stream faster than worse ones, as you can see here (http://forum.doom9.org/showthread.php?t=99402)

[DeathTheSheep]

Um, bond...check the last post ;)...
I used ffdshow with standard, clean x264 decoding

[IgorC]

there are several version of ffdshow for different cpu intel/amd. with sse, sse2....
did you check mplayer, nero decoder ... Nero for my 2 ghz celeron SSE2 is fastest H.264 encoder.

We still wait for your 32 kbps test. You said vp7 is better than x264. Show us.

[IgorC]

And I also said that even if it was surpassed by x264 in SSIM, the 2 files you had looked the same to me with optimal decoding.

no no no . You have told that VP7 is better than x264 for low bitrate. 49 isn´t enough low bitrate for you. Stream bitrate is about 32-450 kbps/s. In many countries dialup 56 kbit/s has the same price or even higher price than ADSL/Cablomodem light 128 kbit/s. And now you're saying that vp7 and x264 are equal. Do you think it's acceptable?

[DeathTheSheep]

I said I have a Celeron 2.93ghz, which supports mmx, isse, sse, sse2. And I used the November 20 ffdshow from x264.nl.

Nero, according to bond's speed test, scored a 50.74fps, which is 2fps slower than the new ffdshow at 52.15fps. According to his particular test, ffdshow is a faster x264 decoder than nero.

We still wait for your 32 kbps test.
If you took the time to read the posts before attacking people, I said:
"I could do it myself....(light bulb goes off)....but I can't download the trailer...."
That's right, I can't download the trailer on my 45k connection. Why do you think I'm concerned about dial-up streaming anyway? Because I have dial-up internet, and I'm not about to lug a 75MB file across it.
You said vp7 is better than x264.
Yeah, thanks a lot. You didn't pay attention to a word I said, so I'm not going to humor you. Sorry. You're beginning to bug me. I clearly said that each person is entitled to his/her own opinion, and my opinion was that VP7 looks about the same as x264 at the bitrate you gave. "About the same" does not mean "better" to me, and if it does to you, I'm sorry. If you're talking about the first post I made about VP7 having better SSIM, my figures were incorrect because they were made with an older version of x264 -- which, I might add, was pointed out in the thread. The same person who reviewed x264 and VP7.0.8's SSIM said that VP7 is better at low bitrates. If you have an issue, take it up with the tester.

Show us.
OK, dude, listen. One last time. I said that it is my [b]opinion[b]. The same clip might look really good to different people and really bad to others. If I were to give you 2 clips, you might judge one as being better than another, and I might pick the opposite clip--that's why there is no "best" codec. That's why it's subjective.

First of all, how can I "show" you something that I think in my own mind based on what I saw before? What, can you somehow leap into my memory and reconstruct my opinions? If you don't think it's better, then don't. We've been over this.

[IgorC]

Deathsheep said yesterday :

VP7 is the highest-ranking public codec in the world right now

Deathsheep said today :

there is no "best" codec


No commentary.

[DeathTheSheep]

And I also said that even if it was surpassed by x264 in SSIM, the 2 files you had looked the same to me with optimal decoding.
no no no. You have told that VP7 is better than x264 for low bitrate.
I got that from my own visual opinion. The test clips you uploaded "looked the same to me" in quality, in my opinion. Do you have a problem with that? Why do you care what I think? If you don't think so, don't--I'm not forcing you.
49 isn´t enough low bitrate for you. Stream bitrate is about 32-450 kbps/s. In many countries dialup 56 kbit/s has the same price or even higher price than ADSL/Cablomodem light 128 kbit/s.
The title of this thread, for those of us who can't read, is "What's The Best Dial-Up Video Codec?"
so I couldn't care less about what they offer in your country. Maybe I would if you weren't so nasty. Try to stream 49kbps on a dial-up connection, which is what this thread is about. Just try it. If you did, you'd see the problem right in front of you. Why don't you leave me be instead of attacking my opinion? You uploaded the samples, people can see for themselves.
And now you're saying that vp7 and x264 are equal. Do you think it's acceptable?
Yes, I think both provide acceptable quality. At first I thought VP7 had higher SSIM, which was based on an older test. Now that you uploaded the files, I can see how much progress x264 made and I can say they are about equal in my opinion based on your sample. We've been over this.


DeathTheSheep said yesterday:
Woah, surpassed already! Wow, the progress x264 is making is truly astounding! The question is: can it be streamed as easily?
Sorry about the misinfo... it was tops until recently.

And sexy ol' DeathTheSheep says exactly the same thing today.

Read the posts. Else you'll sound stupid.

No further comments needed, man

[CruNcher]

Nero, according to bond's speed test, scored a 50.74fps, which is 2fps slower than the new ffdshow at 52.15fps. According to his particular test, ffdshow is a faster x264 decoder than nero.


hmm should be retested the 4.2.1.0 of Neros Decoder beats the 2.2.1.0 ateme version slightly by some fps ;)

[IgorC]

If real dial up speed is aprox 43 kbps.
34-35 kbps video
7-8 kbps audio (speech codec like speex)

Judge for yourself.
http://rapidshare.de/files/8069490/35kbit.zip.html
vp7 isn´t bad but it's far to be best or even on par with x264.

P.S. vp7 2 passes best quality . sharpness 0. sharpness 1 or 2 already blocky for low bitrate.

[DeathTheSheep]

Good, with your test, and x264 does indeed look good.
Speex? With Harry Potter? hmmmmmmmmmmmmm I'd want to hear the sound, so I'd use wma at 8kbps or AAC-plus at 8kbps.
But anyway, I can't reach 43kbps for very long. I tested it and it lasted for about a minute over 43kbps, and levels out at 38kbps max.
Is mine just slower than average? I downloaded your sample file at 5.007 KB/sec average, which is about 40kbps. So it's good 4 streaming, eh?

Anyway, yes, VP7 made a lot of flickering in this sample... What was your keyframe interval?

[ot]
Hey, slightly OT, but does ffdshow20051121 work for you? It wouldn't load the x264 file in a directshowsource avisynth ("video stream not found")...