What specific technical advance(s) allowed PCs to play "Full-screen full-motion" video?

Question

In the early 1990s, the new buzzword for PCs was "Multimedia", and the gold standard for multimedia performance usually talked about was "full-screen full-motion" video playback. Many will remember early digital videos being played in tiny windows on Windows PCs. In contrast, the often-plugged aspiration was to have basically long-playing digital video on the PC at roughly the quality of a VCR / Standard-Definition Television, utilizing the full screen area, around 30 fps, and including realistic sound.

I believe that such capability became commonplace on new PC hardware around the middle of the 1990s. I remember CD-ROM games such as "The 7th Guest" becoming popular based on these sorts of abilities, though that particular game falls short of the quality mark described above.

Naturally, this whole endeavor revolves around hardware bandwidth and codec's, which includes software components. My question is, can it be pin-pointed the specific hardware/software technologies and when they went mainstream on new PC's to support this level of multimedia video?

Answer 1

1. An efficient motion video codec (such as MPEG-1, 1993) to make storage or transmission of digital video practical.
2. Real-time motion video decoder chips such as the C-Cube CL4000 (1993), because efficient codecs are processor-intensive. Or a sufficiently fast CPU such as the Intel Pentium (1993) and a high resolution multimedia timer (I think the necessary circuitry was first incorporated in sound cards before moving to motherboard chipsets).
3. A high speed path between the video decoder chip and the video memory or video output. On PCs, this is achieved either by placing the chip on the video card (such as the ATI Rage 3D, 1996), or as described by @tofro by connecting through a VGA card's high speed feature connector to access video memory directly (such as the Sigma Designs Reelmagic CD lite, 1993), or by switching from ISA to VLB (1992) or PCI (1993).
4. 15-bit color or better (late-'80s), to avoid palette issues.
5. A redistributable high capacity storage medium (such as CD-ROMs (Yellow Book), 1988) or a high speed network (such as satellite Internet, mid-'90s) to get the video data into the PC.
6. Buffered PCM audio such as a Creative Labs SoundBlaster sound card (1989).

To make fullscreen full motion video mainstream, the following technologies were needed:

7. A software abstraction layer for video decoding such as DirectShow (1998).

Answer 2

I'm going to stick my neck out and argue, chronologically:

1. QuickTime;
2. Cinepak;
3. MPEG1.

QuickTime, which launched in 1991 and therefore predates Video for Windows by about a year, introduced a fairly general framework for containers and codecs, and an initial video codec which was a fairly simple thing. No motion compensation, just altering or not altering different 16-pixel parts of the display in each frame, with differing precisions (ranging between a single colour for all 16 to a different colour for each pixel).

Cinepak, from 1992, was the first codec added to both QuickTime and Video for Windows that promised 'full-screen' (i.e. 320x240), 'full-motion' (15fps) video. It's actually broadly similar to QuickTime in just dividing the display into macroblocks and assigning each a particular fill, then modifying those fills frame-by-frame, but there's an extra level of indirection as to block contents, a fixed-size codebook and codebooks receive only sequential updates from one frame to the next, key frames aside.

Cinepak is the codec you're used to from the Saturn, Mega CD, 3DO, etc — everything that produced full-motion video in software.

MPEG-1 was also standardised in 1992 and introduces a substantial number of new ideas:

• use of the frequency domain to describe macroblocks in terms of their significant frequency content (via DCT in this case);
• sampling luminance at a greater frequency than chrominance;
• motion estimation of macroblocks from frame to frame — macroblocks can now be updated from frame to frame not only by content but by position;
• bidirectional frames, which describe a frame not just as a mutation from the prior, but as a combination of mutations from the prior and the next.

That adds up to a substantial extra processing burden, and an extra buffering burden for bidirectional frames but it gets you "VHS quality video" (i.e. 320x240, 30fps) at CD-ROM read speeds, where Cinepak could do only half the frame rate at the same data rate.

A desktop PC could be equipped with something like a ReelMagic card for MPEG-1 playback; consoles like the Philips CDI, Saturn and 3DO tended to have a slot where one could add an after-the-fact MPEG decoder, primarily for playing Video CDs.

A Pentium can decode MPEG-1 in software; a 486-class machine can do something between Cinepak and MPEG-1.

Answer 3

The first real "Multimedia" experience I had was with a Sigma Designs REALMagic MPEG decoder card that was one of the first that could full MPEG decoding in hardware. It had to be connected to the feature connector of your (S)VGA card to transport the picture to the screen. The PC itself didn't have to do much with these cards except maintaining a relatively constant data feed of 2 MBit/s into the card. The first cards of that type showed up in the mid-nineties and replaced rather low-quality software decoding of videos (AVI, Quick Time, Video for Windows)

PCs were far from being capable of doing real-time MPEG-2 decoding in software back then. They had to be supported with a bit of a hardware lift to play proper video.

Answer 4

If you look to unconventional sources, you find some surprising answers.

The Demoscene has demonstrated that full-motion video is actually possible on an XT with a hard drive and a Sound Blaster. Though admittedly, this didn't happen until long after that hardware was all obsolete.

Check out "8088 Corruption" by Trixter/Hornet from 2004 and its follow-up by the same author "8088 Domination" from 2014. The files for both are encoded in formats designed for the job, not your regular h264 codec, though the tools for making them have since been released.

If you don't want to try running them on old hardware or in DOSBox, there are YouTube video links on those pages as well as downloads of the demos themselves. Just note that if you want to try running them on real hardware they are whole-drive-filling huge, at 10 and 30MB.

Answer 5

Another fundamental bit of context beyond the world of PC's is that video cameras were dropping in price, so people were much more likely to be doing video stuff even away from computers so there was suddenly a much larger market for the computer industry to tap into. America's funniest home videos launched in 1989, and wouldn't have been possible without practical portable camcorders being widespread. Likewise, the TV show COPS launched in 1989 because shooting TV on-location without needing a whole video truck was new and impressive, and in 1991 a citizen was able to document the Rodney King beating and it shocked the nation because it was so new for somebody to just have a handy camcorder when something was happening. Such attacks presumably also happened prior to 1991 -- there just weren't as many people with video cameras to document the phenomenon. (Things like portable 8mm film cameras did exist, but the practical differences between film and video are probably out of scope for a computing discussion.)

In 1985, the Amiga launched with fairly decent graphics by the standards of the time, and the ability to use NTSC compatible display timings and genlock so an off the shelf personal computer could be used as a part of TV production for the first time. Avid Media Composer launched in 1989, compatible with a 68020 Macintosh II from 1987, paired with dedicated video hardware. By 1990, the Amiga 2000 with a Video Toaster could do all sorts of live switching, keying, and cheesy digital video effects that used to require exotic dedicated hardware. Apple launched QuickTime in 1991. (I'll get back to that.) In the consumer space, video games were pushing ever more 'lifelike' and 'realistic' graphics as a selling point, kind of culminating in 1992 with Mortal Kombat directly using captured video frames as sprites to make the first video game characters that literally looked 'lifelike.'

So, at the start of the 1990's, you have a general emergence of video related technologies, and a general trend of general purpose personal computers being involved in some aspects of video production. There wasn't any single "lightbulb moment" in that path. Just a whole range of technologies becoming cheaper, more practical, and more desirable in a virtuous feedback loop.

The better the compression ratio of your video codec technology, the less hard drive space and performance you need. The simpler the decoder of your video codec, the less CPU power (or dedicated decoder chips) you need. If you have a video decoder in your display card, you don't need a ton of memory bandwidth between the CPU and the display hardware. If you have a lot of bandwidth, you can use simpler display hardware, etc. There's no one specific thing like, as soon as RAM hit X nanosecond timings, them boom it became possible to do video with that RAM. It was a balance of factors, where if any one technology in home computers had been worse, some other aspect of the system might have gotten more work to compensate.

Decoding a 640x480 24 bit RGB frame of video does require an exact minimum amount of RAM. But that doesn't mean video was impossible with less RAM -- it just meant people used crappier looking lower resolution videos, and videos got a bigger and better looking over time as memory and storage got cheaper. There was too much consumer interest, so nobody was going to throw their hands up and declare it impossible -- they just did they best they could with the constraints of the systems they were targeting at the time.

If I had to pick a single definitive technology that pushed video from exotic to ordinary, I'd say the launch of QuickTime in 1991. At that point you could write video production software while standing on the metaphorical shoulders of giants. You didn't need to implement your own file formats or codecs, and you didn't need to tie your application to specific hardware accelerators. Your customers could buy QuickTime compatible software today, with the reasonable expectation that the crappy video quality would get better over time when they ran your software in the future on better hardware and using better codecs. That meant that people making video could feel safe buying video software tools and not worry about the tool getting abandoned. That confidence in the marketplace resulted in fostering a market for video production related software, which led to a bunch of people being able to make video on their computers, which led to a bunch of people being able to make video for other people to play on their computers in FMV games and whatnot.

Now, QuickTime isn't really a technology, per se. It's just a standardized interface to some hardware and software technology that existed independently of QuickTime. Remember, AVID launched on Mac before QuickTime existed, so people were already working in that space - it absolutely wasn't a hard requirement. But the fact that a major platform vendor was treating video as a major selling point had a huge impact on the market and massively lowered the barriers to entry, so it's what I'd pick as the turning point. It was when everybody noticed that the technology was polished and shiny enough to start the gold rush.