future capture cards

[gibverse]

After reading the Doc’s latest post about real-time MPEG-2 on 600MHz and faster systems, I started thinking on what direction the capture board industry is moving. I truly do think that we are moving towards a world of software codecs, not hardware codecs. And there are advantages to this. I for one like the notion of not being locked into hardware based codecs. I rarely use MJPEG anymore, preferring to use the YUY hack and either MPEG-1 or Divx in real-time for most tasks. And when I need to do certain effects, uncompressed works wonders for short clips.

Although the RT2000 looks really nice, being on a limited budget, doing desktop video purely as a hobby (read getting no money out of it), I decided to scratch the RT2000/DV500 and other such real-time boards for my next system, which I plan to build towards August. At the time, I had just started playing around with MPEG-4/Divx and the GoMotion in MSP 6. It became apparent, that I could actually build a dual-CPU system, with more than enough horsepower to do real-time MPEG, and most other codecs, for less than a semi-pro board. The announcement of MPEG-2 is just the nail in the coffin.

I am not dismissing the merits of these new boards. For those who need real-time effects, they’re a great deal. However, for the amateur enthusiast, a great deal of the users here, I do believe that hardware codecs will become a way of the past. What I wish I knew now is what does Matrox have in store for us next? There is already speculation that the G450 will not work with the Rainbow Runner G-series (see the May 2nd news on matroxusers.com for more).

I personally would be happy with a solution like the following. A PCI card/daughter board like the RR-G, with a breakout box, video out and tuner like the Marvel. Adding an IEEE-1394 port for DV to the BOB would be great as well. And depending on the complexity involved to the design, maybe integrated locked audio, to help with sync problems. And a price between US $100 and $200 would be icing on the cake. Oh, and before I forget, WINDOWS 2000 SUPPORT!

That’s just my thoughts right now. I welcome the rest of the forum to chime in about this, pitch in ideas, or say I’m out of my mind. For all I know maybe there’s an inexpensive solution like this out already. Maybe after some discussion we could create a list of wants and don’t wants which could be sent out to Matrox and other manufactures.

John

[DrMordrid]

At one time I derided the use of softcodecs for capturing. I still do when it comes to certain products with outdated hardware and poor or propriatory codecs, but with the recent changes in processor and codec technology my mind has changed when it comes to many of the new offerings.

First lets make sure we make the distinction between "realtime capture" and "realtime effects".

Realtime effects is still very much a hardware-assisted function and will remain so at least until we hit the 2 ghz generation of processors. As for realtime captures this has been possible with software for some time, albeit at much reduced resolutions and quality.

What has changed is that the new software codecs are becoming very efficient (read: FAST) with quite high quality. This is mostly due to the MMX, 3DNow! and SIMD instructions available on the newer processors and much improved coding in the codecs themselves. This makes them capable of doing realtime D1 encoding (D1=720x480) with cards that can pipe YUV/YUY streams to such a codec.

What is also new with some of the newer softcodecs is support for the OpenDML filespec, which will allow them to capture without any real limitations on file size as long as the filesystem supports this.

Another advantage of the improved softcodecs is the ability to instantly upgrade or even change codecs in your capture system to tailor the capture to a particular task. An example of this would be directly capturing MPEG-1/2 for use in VideoCD and DVD versus capturing and converting. With a fully configurable codec, such as MVideo promises to be, this could be a boon to those activities.

In terms of the capture cards themselves this method should simplify them and reduce costs. As to if these savings will be passed on to the user is up for grabs.

My guess is that some companies will pass on at least part of the savings while others (P*******, etc.) will promote it as a "hot new feature" and charge more for it. Most likely these rogue products will use a propriatory codec and will be unable use standard codecs. Of course alternate compatable codecs might be available for $50 a pop ;-))

Dr. Mordrid


[This message has been edited by DrMordrid (edited 30 May 2000).]

[This_Idiot]

Just to throw some more ideas into the pot. It's commonly claimed that both Linux and BeOS run a lot faster and more efficiently than Windows. Presumably, then, this means that they will become a better platform for supporting soft codec captures ?

[fewmet]

While BeOS has many features that could potentially be useful in digital video, it is not actually very usable for that currently. The only editing app available is something at the level of Videowave. The only company working on a prosumer editor ran out of funds recently (Mediapede). Codec support is not wonderful currently - there's raw RGB, Indeo, some QT formats, MPEG-1 but no MJPEG the last time I checked (there is PJPEG), no MPEG-2 (promised RSN). Performance is not so wonderful on some of these although the ability to maintain playback on multiple windows without falling over on some formats does show its potential superiority over Windows.

The codec API is private so you can't roll your own readily either if you want it to integrate with the rest of the Media Kit. This is ostensibly because the API still remains fluid.

Driver support is limited too. Bt848 capture is supported but Zoran card support wasn't the last time I looked.

With the company's refocusing away from the Media OS angle to the internet appliance, I think we can expect only areas crucial to the latter to receive major attention.

It still remains innovative and promising but you will have to be self-reliant if you want to explore this. And if you are going to all that trouble, you might want to look at AtheOS (www.atheos.cz) too, a GPLed BeOS lookalike written by some BeOS enthusiasts over 7 months. Quite promising considering what they have achieved (GUI, drivers, networking, POSIX, etc) Only runs with a G400 for now!

[Joxx]

The GoMotion clips looks horrible!!

I have a pIII-700 but the GoMotion codec is simply a very compromising codec. I think that we have to wait at least until 3GHz for a non-compromising realtime software codec. But in 3 or 4 years, we will most likely have the new HDTV standard, with higher resolution and so on.

Of course it would be better if we could use a software codec, but cpus will most likely be too slow atleast until late next year!

[DrMordrid]

MVideo is coming soon so there may well be another MPEG choice. I got the data sheet today and it looks pretty good. Up to 720x480, NTSC and PAL, 1 to 6 mbps, frame & field etc, etc.

Dr. Mordrid


[This message has been edited by DrMordrid (edited 31 May 2000).]

[Brian Ellis]

Joxx

I believe HDTV is a myth! It has been talked about but there is still no world standard for it. HDTV camescopes are going to have to grow to get an hour of tape on them (not to mention a Brobdingnagian cost) and NLE, as we know it today, is a no-no, because the hardware would be totally unsuitable, especially Windows-based. It would probably require something approaching a super-computer in processing capacity, with a dedicated operating system/software to provide the required resources.

------------------
Brian (the terrible)

[jeff b]

Someday, I'm hoping that I can have a computer system with full HDTV reception, with an analog I/O port for piping in and piping out, along with 1394 for DV.

Here in the US, I'm still chomping at the bit for a decent PCI card that has a standard tuner and an HDTV tuner, and will display all the various resolutions currently being broadcast on the new digital channels. I can't figure out why the only one available (Hauppage WinTV-D) can only do 480p, and isn't compatible with AGP based systems (what a waste of money!). Pinnacle announced a full res HDTV tuner card a few weeks ago, but who knows when it will finally hit the streets? And, of course, it IS Pinnacle...

It's clear that the market for digital TV in the US just isn't going to move very fast with prices of Hi-Def sets running so high. But the MPEG-2 encoding of HDTV is practically a no-brainer at this point! I mean, this is not a NEW codec, after all...

So, I really think that companies like Matrox, Canopus, et al, are really missing the boat by not running hot onto getting HDTV cards to market. There is an installed base out there for Hi-Def TV reception in the US that's bigger than any NLE market they ever dreamed of! And all they have to do is come up with a freakin' ADD-ON CARD, fer cryin' out loud! I mean, watching Hi-Def at 1080i on a computer monitor will sure beat the hell out of watching it converted to NTSC on a regular TV, no matter how big the screen is...

- Jeff B

[DrMordrid]

MPEG-2 may or may not be the way to go for HDTV. By the time there is a full HDTV standard MPEG-2 may well be supplanted by MPEG-4, which are being touted for broadcast, cable and netcasting. Also relevent is the upcoming MPEG-7 multimedia content description interface standard which can use analog, MPEG or other data streams.

Dr. Mordrid