On2 VP4 Personal Codec

[Pepmanel]

The On2 VP4 codec has been released FREE for personal use only.

Features:
- Can be used with any VFW-compatible video processing utility that accepts third-party codecs (for example, Adobe® Premiere®, VirtualDub, and Cleaner™)
- Comparable quality to MPEG-2 at more than a 75% reduction in bandwidth.
- Near-DVD quality at Broadband data rates. VP4 maintains the highest quality possible while delivering full-screen, full-motion at 640x480 screen resolution, native frame-rates of 24fps for film and 30 fps for video.
- Near-VHS quality at ISDN data rates. VP4 is the only compression technology that can achieve near-VHS quality at a screen size of 320x240 and at 24-30 fps.
- Narrowband capabilities. VP4 can offer video streamed at dial-up rates at 192x144 pixels and 15-20 fps.

Download:
Download On2 VP4 Personal! (http://www.on2.com/vp4.php3)

[Sirber]

Good post!

I'm trying Naruto E38 at 450kbps. VP4 isn't fast but Q seems good. It seems to drop duplicates frames...

[edit]

Quality is BAD BAD BAD. I'll post a preview. Compared to RV9, RV9 is 2-3 times better in visual quality at the same bitrate IMHO. 2-pass mode could help on2 codecs a lot...

http://www.webernic.com/sirber/VP4_VFW.avi

[edit2]
Near-DVD quality at Broadband data rates. VP4 maintains the highest quality possible while delivering full-screen, full-motion at 640x480 screen resolution, native frame-rates of 24fps for film and 30 fps for video. Let me laught! 450kbps is higher than Broadband data rates, and quality is inferior than modern mpeg4/rv9/wmv9. :eek:

[RadicalEd]

broadband is a misleading term :\ my broadband is 768kbps. My dad's is 1.5 mbps. Some people have 384kbps broadband. Sneaky termage :o

[phrentec]

it doesn't seem a lot different than vp3. maybe it'll prove usefull for realtime tv encoding.

[Sirber]

Supposed to be 30% better then VP3. VP5 is 30% better than VP4, and VP6 30% better than VP5. So, VP6 is 60% better than VP4 and 90% better than VP3 :confused:

[superdump]

Not quite. :) If VP3 has a rating of 100, then VP4 would be 130, VP5 would be 195 and VP6 would be 273 (100*1.3*1.5*1.4). Heh heh.

Edit: the website says VP5 is up to 50% better than VP4 and VP6 is up to 40% better than VP5. I amended my data accordingly. :)

[phrentec]

well consider the release as a slight step forward aside from there usually corporate blabber. but then corporate blabber does bring in the big dumb money, i mean AOL would sign with them..

[superdump]

lol @ phrentec

I've just done a quick encode with VP3 and one with VP4 and subjectively compared them. VP4 looks considerably better than VP3. I can't express it as a percentage but the image produced is much sharper and more detailed. It is still however not up to the standard of today's codecs.

[Sirber]

When I saw VP4 I got scared :D. There are free and way better alternative. No 2-pass mod defenitly kill that codec.

[superdump]

Hmm, this makes me wonder if theora can ever be any good seeing that it is based on VP3. :S Are all the VP* codecs evolutionary? By that I mean are they tweaked and altered from the previous version or do they ever start from scratch?

[Sirber]

It's long to start from scratch... I think it's just tweaking and adding/removing algos. Advanced options of VP4 aren't tne same as for VP3.

[superdump]

Then maybe theora can overtake VP6. Just as long as the Xiph guys work on it lots. If they don't then it may as well be dead already. :)

[unmei]

in a quick-and-dirty test vp4 clearly looses against rududu 2003-07-27 (in the 1200-1500 kilobit range).
Now everyone seems to agree this codec is made for much lower bitrates then i am interested in, but i wonder how it could be good at 450 or so kilobit if it looks already at 1200 kilobit like a 450 kilobit rv9 ???
well i will test mid-low bitrates but i feel it won't change much of my first impression and i better invested more time in rududu testing.

Now lets hope the xiph people are more effective in tweaking than on2 - i would hate Theora looking like vp-what was it again -2? -4? -7? :)


i missed the "temporal resample" switch. it seems to help increase compression a lot. I also realised i was at the outmaxing limit of the codec, which is a shame: it means codec cannot produce hi quality pictures at all. It even seems me like the codec doesn't gain anything above 800 kilobit (for my sample - the limit might differ but the fact that there is no good quality remains).

And well lol 480 kilobit: i don't need to check this on the TFT, i see on this cheap CRT in normal play how poor it performs. i never tested this but i would bet rv9 can do the same at 250-300 and even xvid (which i consider a hi quality codec particularly not-suited for low rates) might be "better" at 480 - if you can compare pixelstorm to bumping watercolor version :s

Also i observed frame size graph in VDM looks like in CBR encode as soon as the data rate is a bit higher (not for the keyframes, but for predicted ones), probably because the codec thinks it cannot give more quality anyway.

Oh someone stop me or i will bash this codec all the morning ;)

[Sirber]

Go ahead. I'm already basing it since my first post :D. I don't know why on2 released it, it's bringing shame on them :eek:

[Tommy Carrot]

Back at that time, when they did this codec, they told in the press releases, VP4 is superior to mpeg4. :D

Hell, rududu beats VP4 comfortably at every bitrate, and it's far from mature.

VP4 not only sucks because of the missing 2-pass function, but the compression efficiency is far from today's standards.

Now i understand why they did not released this codec until now. Their press releases and the codec are very far from each other.

[Sirber]

They should pay us for installing it :D I lost my time installing and testing it :devil:

[Phanton_13]

in reality the true problem is with ON2 political in the developement of a codec, they develop the codecs very fast( they change algoritms, specifications and tecniques, in other words they change the aproximation to the compresion problems very fast) and don´t mature their existing codecs in other words they codecs are in reality unoptimized, intestead of this the competenze are optimized their algoritms, you can compare Xvid against proyect mayo(divx4), if you use a compatibility profyle in xvid, you can see the real superiority of the mature xvid, can you imagine this evolution in vp3 or vp4??

[Tommy Carrot]

Yeah, but they have the freedom of using any algorithms. So why finetune, if there are better methods? They have easier job to create a decent codec, they have no restrictions.

[Sirber]

The codecs seems also CBR. Is this codec supposed to be for streaming?

[Phanton_13]

Yes they have this freedom but not is the change of the algorithms the matter, is the mode than they made it. Is better if a company take out 3 diferenters codecs or they only take out 1 codec and optimized this( on2 vs mepg4(proyect mayo + Xvid/divx5))?

[midiguy]

Thinking back to the samples I watched, VP5 wasn't really better than VP4 (maybe slightly better), but VP6 is definately worse than the whole lot. Go watch the samples on their website! Gross!

[Sirber]

I don't want to install their crappy player. I tryed one time "Beautifull Mind". I taught: This is great for 300kbps! But, it was at 1000kbps :D Every codec can beat it's quality at 1000kbps, even DIVX3 SBC.

[deXtoRious]

You could compare a DivX3 SBC video at 450 kbit/s with a 1200 kbit/s VP4 (maybe) and that would be a fair test. Otherwise don't bother - DivX3 is WAY better.

[Sirber]

What about MPEG4v1 versus VP4? :D

[deXtoRious]

Better, but still... let's take a 900 kbit/s MPEG4v1 versus 1200 kbit/s VP4, okay? :D