25 Years Ago: DivX – The MP3 for Video is Born

25 Years Ago: DivX – The MP3 for Video is Born

In the era of analog technology, once-defined standards like PAL color television and FM stereo lasted for decades. Even digital technologies can have long-lasting popularity, as evidenced by the Compact Disc being over 40 years old and the MP3 audio compression method nearing its 30th birthday, both of which remain widely used today.

Digital video, however, operates on a different timeline. Starting in the mid-1980s, the Moving Picture Experts Group (MPEG) has been developing methods to compress the massive data flow of digital TV images to manageable sizes. Without compression, video in the outdated standard definition requires around 270 Mbps, and Full HD needs at least 1 Gbps. Initially, only studio video recorders costing several hundred thousand Deutsche Marks, a sum equivalent to today’s euros, could handle this data flood. These devices were never practical for the average consumer.

MPEG has always followed the right principles for lossy compression of digital video. The first practical implementation, known as MPEG-1, gained somewhat dubious fame in the early 1990s on Video CDs. The parameters used for these early video discs promised VHS-quality on paper. However, in practice, the jerky, block artifact-ridden images disproved any such optimism. But with higher bit rates, full resolution of standard video formats like PAL and NTSC, and the ability to process interlacing, MPEG-1’s fundamentals—as MPEG-2—took off worldwide in the mid-1990s. It was first adopted for Digital TV and from 1996 for Video DVDs, which became a global success. With MPEG-2, a mere 5 to 7 Mbps was sufficient for a decent image instead of the aforementioned 270 Mbps.

Why recount this history? By the turn of the millennium, the technical foundation of DVD video encoding was old news, with newer, more efficient codecs available that could have enabled better images and/or longer playing times on DVDs. However, when work on the Video DVD and corresponding playback devices started, it was too costly and complex to implement these more efficient codecs in hardware. DVD players with MPEG-2 support available in Europe in 1997 cost about 1500 Deutsche Marks, a price that would have likely exceeded the 2000 Deutsche Mark threshold with a more efficient codec. MPEG-2 seemed an acceptable compromise between quality and cost.

While manufacturers, engineers, and program providers dwelled on details like copy protection and the ideal audio format for these novel discs, video codecs improved, and PCs and dedicated hardware became faster and cheaper. After MPEG-2 naturally came MPEG-4, which, in a usual MPEG-confusing fashion, includes several related but not necessarily compatible video codecs. While only H.262 is permitted for MPEG-2 as a video codec, MPEG-4 includes three distinct codecs, H.263, H.264 (“AVC,” standard for HDTV in Europe), and H.265 (used in DVB-T2 HD and various modern video cameras).

Microsoft began working with audio and video codecs in the mid-1990s, hoping to capitalize with its own version of MPEG-4. Users first encountered its related codec in the .asf (Advanced Systems Format), later known as .wmv (Windows Media Video). These MPEG-4 Version 3 libraries distributed by Microsoft as “Media Tools 4” did not produce ISO-compliant MPEG-4 video. Microsoft’s alterations to the Windows Media Player with these tools led to it being unable to play demo videos by French computer artist Jérôme Rota (nickname: “Gej,” meaning crazy in Occitan language). This led Rota and a German hacker with the alias Max Morice to extract the Microsoft codec, hack the library to support higher bit rates, and enable its use with the de facto standard video container for computers, AVI (Audio Video Interleave), since the Media Tools were constrained to Microsoft’s proprietary ASF. Rota released this hacked, unleashed codec as “DivX;-) 3.11 Alpha.” Later, with three other hackers, he founded “Project Mayo,” reportedly chosen because “DivX and mayonnaise are both French and difficult to produce.” The name DivX itself is a play on a failed US DVD rental system called Digital Video Express.

The early hacked codec wasn’t truly universal. Rota and his colleague noticed that version 4.1.00.3290 of the MS codec was better for encoding slow scenes, while the older version 4.1.00.3917 excelled with fast motion. They built a “Low Motion” variant (FourCC code “DIV3”) and a “DivX;-) Fast Motion” version (FourCC “DIV4”). Decoders could handle videos created with both variants, but users had to choose the appropriate variant for their source material. This didn’t change with the subsequent DivX;-) 3.22 release (also known as DivX;-) 3.11 VKI), which now detected scene changes and automatically inserted a keyframe, a full picture—one of many methods for video data reduction.

Early video codecs used on PCs overlooked two important details for SD video: PCs and digital cameras worked with square pixels from the outset, and to maintain the traditional 4:3 aspect ratio (equivalent to 1.33), the resolution would be 640×480 pixels. Conversely, for digital SD-TV, the resolution aligned with the active line raster of analog NTSC or PAL, using 486 or 576 pixels, respectively, per line and 720 pixels within each line. With square pixels, this translated to an aspect ratio of 1.48 or 1.25, not the familiar 1.33. Therefore, digital SD video used rectangular pixels, and to view such an image on a PC monitor correctly, the playback software needed to stretch or squash the image to the appropriate width, either 640 or 768 pixels.

Further complicating matters, from the 1990s, the transition to the more cinema-like and human-vision-friendly 16:9 aspect ratio (1:1.78) began. For SD digital video, the resolution stayed at 720×480 or 720×576 pixels. Using an anamorphic recording trick borrowed from the cinema, the wide image was compressed, resulting in an egg-shaped image when displayed on a 4:3 TV. A 16:9 TV, however, stretched the image back to the correct proportions. Early codecs and player software were unaware of this technique—those desiring to view wide images on a PC screen had to supply more pixels, resulting in resolutions of 854×480 or 1024×576 pixels
Read More