Updated November 2026: YouTube’s early-2027 AV2 plan, the first AV2 hardware decoder IP, HDMI 2.2 and Ericsson’s 2026 traffic figures, plus answers on video in a computer and digital footage.

What is digital video?

Digital video is moving pictures stored as numbers: a sequence of still frames, each a grid of pixels, and each pixel a small group of values a computer can store, copy, compress and send. Analog video records the picture as a continuous electrical wave instead, so every copy and every transmission adds noise.

Every video you stream, download or record on a phone today is digital. Five numbers describe it (resolution, frame rate, colour space, bit depth and scan type), and a sixth, bitrate, decides what it costs to deliver.

In computer terms, a video is a file or a live stream that holds those numbered frames, usually compressed by a codec and wrapped in a container such as MP4, along with the audio that plays with them. Digital footage is the same thing from the camera side: the clips a digital camera, phone or webcam records as files before anyone edits them.

Analog video Digital video
How the picture is stored A continuous electrical wave Discrete numbers: frames → pixels → values
Copies Each generation loses quality Bit-for-bit identical
Compression Practically none Codecs shrink it hundreds of times (H.264, HEVC, AV1)
Software processing Not possible without converting first Editing, search, AI analysis on any computer
Typical examples NTSC/PAL broadcast, VHS tape Streaming, Blu-ray, video calls, IP cameras

Why this matters

If you are about to build a product that uses video, whether a Netflix-style service, an online doctor visit, a video conferencing app, a security camera platform or an online learning tool, expect a year of conversations about video that sound like another language. Engineers will tell you that you need "H.265 at 4K with 4:2:0 chroma and 10-bit depth", and you will have to decide whether that fits your customers, your budget and your timeline. Video is also most of what networks carry: around 75% of all mobile data traffic at the end of 2025, according to Ericsson, and total mobile traffic passed 220 EB a month in Q2 2026, up 23% in a year. 11 We have built more than 250 video products at Fora Soft since 2005, from concert streaming for 10,000 simultaneous viewers to AI video surveillance, and this is the primer we walk every new client through. By the end you'll be able to read a vendor's pricing page, a YouTube engineering blog post or an FFmpeg command and see the trade-offs in it.

We'll build the picture piece by piece, and you don't need any prior knowledge. Every term gets defined before we use it, and the important ones come back every few sections so you don't have to hold everything in your head at once.

From analog to digital: a 90-second history

For most of the 20th century, television and video were analog. Here "analog" just means continuous: a wave of electricity that varied smoothly with the changing brightness of a scene. A camera turned the light it saw into a wavy electrical signal and broadcast it over the air. A television set used that signal to steer a beam of electrons across a glass screen, painting the picture line by line at high speed. There were no numbers anywhere in the chain. The picture was a wave, and the screen drew the wave.

Three regional standards ruled the analog era. They differed in how many horizontal lines they used and how fast they refreshed:

  • NTSC: North America and Japan. 525 lines per picture, 60 refreshes per second. Adopted in black-and-white in 1941 and updated for colour in 1953. 1
  • PAL: most of Europe. 625 lines, 50 refreshes per second. Patented by the German company Telefunken in December 1962. 2
  • SECAM: France and parts of the former Soviet bloc. Same line count as PAL, with a different way of encoding colour.

The analog system worked, within limits. Tapes degraded every time you copied them. The signal carried only so much detail before the picture went fuzzy. And computers couldn't do much with it: you couldn't process it, search it or send it efficiently over the internet, because computers think in numbers and the signal was a wave.

The shift to digital video ran from the mid-1980s to 2009. Digital means sampled into numbers: in place of a continuous wave you get a long list of discrete values that a computer can store, copy, process and send. Studios went first. Sony's D1 recorder, launched in 1986, stored uncompressed standard-definition video as numbers on tape. 12 Consumer milestones came quickly after: MPEG-1 (the first widely used digital video format) in 1993, DVD-Video in 1996, the ATSC standard for digital terrestrial television adopted in the United States in 1996, and similar standards across Europe under the name DVB. By June 2009 US law required all full-power over-the-air TV broadcasters to switch off their analog transmitters (low-power stations followed in July 2021). 3 Every camera, disc and stream you and your customers touch today is digital.

Digital won for three reasons that still shape any product you build. Digital copies are perfect, because copying a number doesn't degrade it the way copying a tape did. Digital can be compressed: you throw away information the eye can't see and shrink the file enormously, which is what makes internet streaming possible at all. And software can work on it, so you can apply filters, run AI, search it or edit it on any laptop. You can do none of that with a continuous wave.

What's actually inside a digital video file

At its core a digital video file is simple. It's a long ordered list of still pictures called frames, played back at a fixed speed so your eye sees motion instead of slides. A flipbook works the same way: a stack of slightly different drawings, flipped fast enough that the brain stitches them into movement. A video file just has thousands of frames per minute, and each one carries far more detail than a flipbook drawing.

Each frame is a flat, two-dimensional grid of tiny coloured dots called pixels (short for picture elements). Picture a bathroom floor covered in very small mosaic tiles, each a single solid colour. From far away you see a smooth picture; up close you see individual squares. A digital image works exactly like that. A Full HD frame, for example, is a grid 1,920 pixels wide and 1,080 pixels tall, just over two million tiles per frame.

Each pixel is stored as a small group of numbers. The most intuitive way to store colour is three numbers, one each for red, green and blue, because mixing those primaries in different proportions produces a very wide range of colours (not every colour the eye can see, as the colour-space section shows). This system is called RGB, short for Red-Green-Blue, and it's how monitors and cameras handle colour. A bright red pixel might be stored as (255, 0, 0), a bright blue one as (0, 0, 255), and a grey one as (128, 128, 128), with equal amounts of each colour.

Here is the first counter-intuitive thing about video. Almost every video codec, the software that compresses video for storage and streaming, does not work in RGB. It converts each pixel to a different three-number system called YCbCr. In YCbCr the first number (Y) is the luma, meaning brightness only, how light or dark the pixel is. The other two numbers (Cb and Cr) describe the chroma, meaning colour information only: how blue and how red the pixel is, apart from its brightness.

The reason is a quirk of the human eye that video engineers love. Your eye is far more sensitive to changes in brightness than to changes in colour. You'll spot a bright detail going fuzzy at once, but a colour going slightly fuzzy slips past, especially in a moving image. Once brightness and colour are separate, a codec can spend plenty of bits on brightness and drop most of the colour bits, and people still see a great picture. That trick alone roughly halves the size of a video file before "real" compression starts. Its technical name is chroma subsampling, and the colour space article covers it in depth. More on YCbCr and colour spaces here.

Three zoom levels of a digital video file: a stack of frames, one frame as a grid of pixels, and one pixel as the three numbers Y, Cb and Cr Figure 1. Three zoom levels of the same video. A file is a sequence of frames. A frame is a grid of pixels. A pixel is three numbers: one for brightness (Y, luma) and two for colour (Cb, Cr, chroma). Codecs work in YCbCr because the eye cares more about brightness than colour.

That structure (file ➜ list of frames ➜ grid of pixels ➜ trio of numbers per pixel) is the whole mental model. The rest of this article keeps asking one question in different ways: "how many numbers do we use, and what do they mean?"

The five things that describe any video

Every spec sheet, camera setting and streaming app config comes down to five values. Get these five right and the rest of the technology stack follows.

Resolution — how many pixels per frame

Resolution counts the pixels in one frame. The two numbers you'll see, for example 1920 × 1080, are the width and height of the pixel grid. Multiply them for the total pixel count per frame: 1920 × 1080 ≈ 2.07 million pixels per frame. Names like "Full HD" or "1080p" are shorthand for the same thing. 4

The standards in active use in 2026:

Common name Pixel dimensions Total pixels per frame Where you meet it
SD 480p 720 × 480 345,600 Legacy broadcast, old surveillance archives, DVDs
HD 720p 1,280 × 720 921,600 Video conferencing, mobile streams
Full HD 1080p 1,920 × 1,080 2,073,600 Mass-market Netflix-style services, online learning, telemedicine
4K UHD 3,840 × 2,160 8,294,400 Premium streaming, modern TVs (cinema 4K is slightly wider: 4,096 × 2,160)
8K 7,680 × 4,320 33,177,600 Flagship demos, niche broadcast

One detail in those numbers matters more than the rest. When you double each side, you quadruple the pixel count. 1080p has roughly 2 million pixels. 4K is twice as wide and twice as tall, so the total jumps to about 8 million, four times as many. 8K is four times as many again, about 33 million. So each step up in resolution costs far more bandwidth, storage and processing power than it sounds like: the data roughly quadruples.

Frame rate — how many pictures per second

Frame rate is the number of frames shown per second, written as fps (frames per second). 5 You'll meet three families of frame rates:

  • Cinematic: 24 fps. The standard for movies since 1927. Hollywood settled on 24 fps because it was the slowest rate at which synchronised sound still played back smoothly on mechanical film projectors, and because film stock was expensive. 6 Audiences came to link the slight blur of 24 fps with a "movie feel", and the standard has held for nearly a century. Almost every feature film you've seen was shot at 24 fps.
  • Broadcast: 25 or 30 fps. Television and most YouTube videos use 25 fps in PAL regions (Europe) and 29.97 or 30 fps in NTSC regions (US and Japan). The split dates from the 1940s and 50s, when TV refresh rates were tied to the local mains frequency: 50 Hz in Europe gives 25 frames, 60 Hz in the US gives 30.
  • High motion: 50, 60, 120, 240 fps. Used for live sports, gaming streams, action footage and virtual reality. Higher frame rates make fast motion look smoother and show more detail in moving objects.

The trade-off: a higher frame rate means more data per second. Going from 30 fps to 60 fps doubles the number of frames the system has to capture, encode, store and deliver. It also makes fast motion look much smoother, which is why sports broadcasters pay for it.

Colour space — which red, green, and blue?

Colour space is a definition that pins down exactly which red, green and blue (and every shade in between) the numbers in your video refer to. That sounds odd, since red seems to be just red, but it isn't. Different display technologies can show different ranges of colour, so your numbers have to say which range you mean, or the picture will look wrong on the other end.

Compare it to the word "warm". Ask a Norwegian and a Saudi each to bring a "warm" jacket and you'll get very different jackets, because their everyday reference for "warm" differs. Colour spaces fix the reference points in the same way.

You'll meet three colour spaces in 2026: 7

  • BT.709 for High Definition (1080p) video. Released by the international standards body ITU in 1990. Covers about 36% of the colours a human eye can see.
  • DCI-P3 for digital cinema and many modern phones and laptops. Covers about 54% of human-visible colour, with much more vivid reds and greens than BT.709.
  • BT.2020 for 4K and HDR (High Dynamic Range) video. Covers about 76% of human-visible colour. HDR video is mastered in BT.2020 colours (with the PQ or HLG brightness curves defined in ITU-R BT.2100); today's best TVs reproduce most of DCI-P3 and part of BT.2020.

You don't need to memorise the percentages. A bigger, newer colour space can make the picture more vivid, but only on a screen able to display that colour space. Mismatched colour spaces (say, a BT.2020 file shown on a BT.709 screen with no conversion) make everything look washed out or wildly oversaturated. More on colour spaces.

Bit depth — how precisely each colour number is stored

Bit depth is the number of bits used to store one colour number for one pixel. More bits give a more precise colour gradient.

A quick refresher: a bit is the smallest unit of computer storage, either a 0 or a 1. With 8 bits you can store 2⁸ = 256 different values (every extra bit doubles the number of combinations). With 10 bits you can store 2¹⁰ = 1,024 values. With 12 bits, 2¹² = 4,096 values.

Consumer video has traditionally used 8-bit depth: each of the three colour components of a pixel takes a number from 0 to 255. So each pixel can be any of 256 × 256 × 256 ≈ 16.7 million distinct colours, which is enough for most everyday content. HDR (High Dynamic Range) video needs much finer steps, because it has to show very bright highlights and very dark shadows in the same scene without visible steps in smooth gradients. That problem is called banding: instead of a smooth sky you see thin stripes of slightly different blues. HDR therefore uses 10-bit or 12-bit depth, 1,024 or 4,096 shades per channel, or roughly a billion to 68 billion distinct colours in total. More on bit depth and banding.

In concrete terms, 10-bit video carries about 25% more data than 8-bit, and 12-bit about 50% more. Worth it for premium content, overkill for a security camera.

Scan type — how each frame is drawn

Scan type describes how the lines of each frame are drawn. There are two options.

Progressive scan draws every line of every frame in one pass, top to bottom. Each frame is one complete, coherent snapshot of the scene. Almost everything modern uses it: every Netflix stream, Zoom call and YouTube video, and every modern TV.

Interlaced scan is a trick from the 1930s for fitting motion into the limited bandwidth of analog broadcast TV. It splits each frame into two halves: the odd-numbered horizontal lines first, then the even-numbered lines a fraction of a second later. The TV stitches them together so quickly that the eye sees one whole picture. The catch is that anything moving between the two halves leaves a visible saw-tooth pattern at the edges, called combing. Modern displays no longer have an electron beam sweeping the screen, so interlaced video doesn't even display properly on them without extra processing. More on progressive vs interlaced.

For any product you ship in 2026, treat interlaced as a legacy input format you might receive from old archives or some broadcast feeds, and never create it yourself.

The bitrate problem — why raw video is impossibly large

This is where "video is just numbers" becomes a concrete engineering problem. Raw, uncompressed digital video is so large it would be unusable on any real network. The maths is short and worth walking through, because every codec decision you'll ever make is about managing this number.

Bitrate is the number of bits a video uses per second of playback, and it's the most important single number in the field. It's measured in bits per second (bps), or in thousands (Kbps), millions (Mbps) or billions (Gbps) of bits per second. Think of bitrate as the width of the pipe your video flows through, like water pressure in a hose. A low-quality video gets by with a thin pipe; a high-quality one needs a much thicker one.

The formula for raw, uncompressed bitrate is:

bitrate = width × height × bits_per_pixel × frame_rate

Take an everyday Full HD video: 1920 × 1080, 30 frames per second, 8-bit colour, with the 4:2:0 chroma subsampling mentioned earlier. bits_per_pixel works out to about 12 bits per pixel on average (the long story is in the colour articles; the short version is that 4:2:0 gives you 1.5 bytes per pixel instead of 3). Step by step:

bitrate = 1920 × 1080 × 1.5 bytes × 30 frames
        = 1920 × 1080 × 1.5 × 30
        = 93,312,000 bytes per second
        ≈ 93 MB/s
        ≈ 746 Mbps

That's 746 megabits per second for raw Full HD. For scale, Singapore, the country with the fastest home broadband in the world, had a median fixed download speed of about 462 Mbps in August 2026 (Ookla Speedtest Global Index) 13, well short of a single uncompressed 1080p stream. A 90-minute movie at that rate would take about 504 gigabytes, more than ten 50 GB dual-layer Blu-ray discs for one film.

Now the same maths for 4K at 60 fps with 10-bit colour:

bitrate = 3840 × 2160 × 1.875 bytes × 60 frames
        ≈ 933 MB/s
        ≈ 7.46 Gbps

7.46 gigabits per second. That's a fire hose, and almost no consumer network on earth can handle it.

A cable inside a building can carry it; the internet between buildings cannot. HDMI 2.2, released by the HDMI Forum in June 2025, offers up to 96 Gbps, so a set-top box can send uncompressed 4K60 to a TV, and broadcast studios move uncompressed video over SDI cables or dedicated IP networks. 17 Once video has to travel over the internet, it has to be compressed.

Now compare that with what people actually receive in 2026. YouTube recommends uploading 1080p SDR at 8 Mbps (12 Mbps at 48–60 fps) and 4K SDR at 35–45 Mbps (53–68 Mbps at 48–60 fps). 8 Netflix asks viewers for at least 15 Mbps to stream Ultra HD 4K and 5 Mbps for Full HD. 9 Raw 4K at the 24 frames per second most films use, with 10-bit colour, is about 2,990 Mbps, so Netflix's 15 Mbps is roughly 200 times smaller; against the 4K60 figure above the gap is nearly 500 times. That ratio is why video codecs exist, and why every streaming business ends up choosing and tuning codecs for a living. More on the math behind bitrate.

Bar chart on a logarithmic scale comparing one 90-minute 1080p film stored uncompressed at 504 GB, as a ProRes 422 HQ master at 50 GB, and as H.264 at 17 GB and 3.4 GB Figure 2. Compression turns a 504 GB raw film into a 3.4 GB streamable file, about 150× smaller, with no quality loss most viewers can see.

Codecs and containers — the confusion almost everyone has

The most common mix-up in product conversations is treating MP4 and H.264 as the same thing. They are different, and the difference matters because devices support codecs, while the container is only the wrapper.

A codec is software that compresses raw frames into a compact stream of bits and decompresses them on the other end. The word is short for coder-decoder. Modern codecs include H.264, H.265 (also called HEVC), VP9 and AV1. Each is a set of mathematical recipes for shrinking video while keeping quality. International standards bodies and industry consortiums release a new one every few years, and each generation is roughly 30–50% more efficient than the last.

Where things stand in 2026: H.264 is still the codec almost every device can play, and YouTube still asks for H.264 High Profile, progressive scan and 4:2:0 chroma in an MP4 container for uploads. 8 AV1 is catching up fast and carried about 30% of Netflix viewing by December 2025. 14 The next generation is already out: the Alliance for Open Media released AV2 in June 2026, 15 and the rest of the chain is lining up behind it. Chips&Media finished the first AV2 hardware decoder IP for chipmakers in May 2026, 18 VideoLAN announced dav2d, an open-source AV2 software decoder, in July 2026 to cover the wait for hardware, 19 and YouTube says AV2 streams will reach its web player as soon as early 2027. 20 By March 2026 more than 90% of YouTube’s catalogue, weighted by watch time, was already available in AV1. 20 What AV2 changes and when to adopt it.

A container is a file format that wraps the compressed video stream, the compressed audio stream, subtitles, chapter markers and metadata into a single file. Common containers are MP4, MKV, WebM and MOV. A container compresses nothing; it only defines how the streams are organised inside the file.

Think of the container as the cardboard box and the codec as the language of the book inside it. The same box (MP4) can hold a book in any language (H.264, H.265, AV1). The same book (H.264) can be packed in several boxes (MP4, MKV, MPEG-TS), depending on whether you're streaming it, archiving it or broadcasting it. 10

Here is why it matters to you. When you tell a developer "ship the video as MP4," you haven't said anything specific about quality, file size or compatibility. Their next question will be "MP4 with what codec inside?", because Safari will happily open an MP4 file and then refuse to play it if the codec inside is, say, AV1 and the user's device has no AV1 decoder. Always specify both. The right sentence sounds like "MP4 container with H.264 High Profile, level 4.1 inside." More on containers.

How video gets from a camera to a viewer's screen

Every video product runs the same six-stage pipeline, from a Netflix-style streaming service serving 50 million viewers to a one-doctor telemedicine call.

The six-stage video pipeline - capture, encode, package, deliver, play and display - with the encoding and decoding stages highlighted Figure 3. The end-to-end video pipeline. The same six stages run inside Netflix, Zoom, a video doorbell, and an online doctor visit. Only the latency budget and the protocol change.

1. Capture. A camera sensor turns light into a digital signal. The sensor is a grid of tiny light-sensitive cells; each cell measures how much light hits it during the exposure for one frame, and that measurement becomes the pixel values. This stage outputs raw frames, the giant uncompressed numbers we calculated above.

2. Encoding. A codec compresses those raw frames into a small stream of bits. It throws away information your eye can't see and finds patterns that repeat within a frame and between frames. This is where the hundreds-fold shrinkage happens: about 150× for a typical 1080p film, about 200× for a 4K one.

3. Packaging. Software called a packager takes the encoded stream, wraps it in a container format and chops it into short segments (usually 2–10 seconds each). Short segments are what make adaptive streaming protocols like HLS and MPEG-DASH work: the player grabs one segment at a time and can switch quality between segments without anyone noticing.

4. Distribution. A Content Delivery Network (CDN) is a worldwide network of servers that keeps copies of those segments close to viewers. When you press play in Spain, the segments come from a server in Spain rather than a data centre in California, which keeps the video smooth and fast.

5. Playback. The viewer's app or browser downloads segments, runs the codec in reverse to decode the bits back into frames, and hands the frames to the display.

6. Display. The screen (phone, laptop or TV) turns the frames back into light the viewer sees.

Different products run different latency budgets through this pipeline. Netflix-style on-demand can allow minutes between filming and viewing, because nobody is watching live. A live sports broadcast shrinks the budget to seconds: standard HLS and DASH lag well behind real time, and their low-latency modes bring that down to a few seconds. A Zoom call or a telemedicine consultation shrinks it again to a few hundred milliseconds. ITU-T Recommendation G.114 advises keeping one-way delay under 150 ms for natural conversation and treats 400 ms as the upper limit. 16 The pipeline stays the same while the rules around it change.

Common mistake: assuming higher resolution always looks better

A 4K stream at 5 Mbps almost always looks worse than a 1080p stream at the same 5 Mbps, because the same data budget is spread across four times as many pixels. Each pixel gets less data, so each pixel is fuzzier, and the picture fills with compression artefacts.

Judge quality by "bits per pixel" rather than by resolution. A 1080p stream at 5 Mbps has roughly 2.4 bits per pixel per second. A 4K stream at the same 5 Mbps has only about 0.6 bits per pixel per second, four times worse. Resolution delivers what it promises only if the bitrate scales with it. So for any video product, skip "is it 4K?" and ask "is it 4K at a bitrate that supports 4K?"

Frequently asked questions

What is digital video in simple words?

Digital video is a list of still pictures stored as numbers and played fast enough (usually 24 to 60 pictures per second) that your eye sees motion. Each picture is a grid of pixels, and each pixel is three numbers for brightness and colour.

What does digital video mean compared with analog?

Digital video is sampled into discrete numbers; analog video is a continuous electrical wave. Digital copies are perfect and can be compressed by codecs and processed by software, while analog copies lose quality every time.

Are video productions still transmitted using analog signals today?

No. Production, broadcast and streaming are all digital today. The United States switched off full-power analog TV broadcasting on 12 June 2009, and every camera, disc and stream in use since then is digital. 3

Is a video you download or stream from the internet digital?

Yes. Anything you download or stream, whether an MP4 file, a YouTube or Netflix stream or a video call, is digital video, compressed by a codec such as H.264, HEVC or AV1 and wrapped in a container such as MP4.

What is bitrate in digital video?

Bitrate is the number of bits a video uses per second of playback, measured in Mbps. Raw 1080p at 30 fps needs about 746 Mbps; after compression, YouTube recommends uploading it at 8 Mbps.

What is the difference between a codec and a container?

A codec (H.264, AV1) compresses the pictures; a container (MP4, MKV) only packages the compressed video, audio and subtitles into one file. "MP4" on its own does not tell you which codec is inside.

What is video in a computer?

A video in a computer is a file or stream of still frames stored as numbers and played back at a fixed rate, such as 30 frames per second. The computer decodes the compressed frames with a codec (H.264, HEVC or AV1) and draws them on screen in order.

What does digital footage mean?

Digital footage is video recorded by a digital camera, phone or webcam and stored as files rather than on film or tape. “Footage” usually means the raw clips before editing.

What are the characteristics of digital video?

Digital video is stored in binary, as distinct numerical values for every pixel of every frame, so it can be copied without loss, compressed by codecs and edited on any computer. Video on tape or a film reel is analog.

How does video streaming work?

Streaming plays a video while it is still downloading. The server cuts the compressed video into short segments of 2–10 seconds, and the player fetches, decodes and shows them one after another without waiting for the whole file.

When did digital video replace analog?

Studios moved first with Sony’s D1 recorder in 1986, DVD-Video arrived in 1996, and the United States switched off full-power analog TV broadcasting on 12 June 2009. Since then cameras, discs, broadcasts and streams have all been digital.

Does 4K always look better than 1080p?

Only when the bitrate grows with it. 4K has four times the pixels of 1080p, so at the same 5 Mbps each pixel gets a quarter of the data, and the picture usually looks worse.

↓ Free download: HDR Readiness ChecklistOne-page A4 audit of an HDR pipeline from capture to display — 21 checks plus the two most common production pitfalls.

Where Fora Soft fits in

We have been writing video software at Fora Soft since 2005, and we work inside the pipeline above every day. If your product has to turn camera frames into a stream people can watch, our video and audio streaming software development team picks the resolution, bitrate ladder, codec and protocol for your audience and builds the pipeline end to end. For real-time products we build on WebRTC: video conferencing platforms with sub-second latency. For cameras we build video surveillance and VMS software. The same five parameters and six pipeline stages constrain every one of those products, and the online learning, telemedicine and AR/VR products we ship too. What changes is the latency budget, the codec choice, the streaming protocol and the device matrix. Every new engineer on our team is expected to know this article's material, and it's the first thing we walk a non-technical product team through when a project starts. For the business view of the same stack (how digital video works end to end, with real CDN cost maths), read our buyer's guide on the blog.

Video Encoding: A Complete Guide to Video Compression - book cover

The book

Video Encoding: A Complete Guide to Video Compression

This article is the first step of a much longer road. The book walks the same plain-language path across 1,500+ pages — from pixels and colour to H.264, HEVC and AV1, rate control, quality metrics, and where AI actually helps. Written by Nikolay Sapunov, CEO at Fora Soft, out of 20 years of shipping video products.

Read it on Amazon →Kindle & paperback · Fora Soft Video Engineering Handbook

Building a product that has to move video?

Tell us what you are building and we will tell you what it takes — resolution, bitrate, codec, latency budget, and the pipeline behind them. 250+ video products shipped since 2005.

Book a 30-min call →WhatsApp →Email us →

See our case studies


  1. NTSC standard history and 1953 colour update. Wikipedia: NTSC, accessed May 2026. https://en.wikipedia.org/wiki/NTSC ↩

  2. PAL standard, patented by Telefunken in December 1962. Wikipedia: PAL, accessed May 2026. https://en.wikipedia.org/wiki/PAL ↩

  3. U.S. full-power analog shutoff on June 12, 2009 (low-power and Class A stations on July 13, 2021); ATSC adopted 1996. Wikipedia: Digital television transition, accessed May 2026. https://en.wikipedia.org/wiki/Digital_television_transition ↩

  4. ITU-R BT.709 parameter values for HDTV. International Telecommunication Union. https://www.itu.int/rec/R-REC-BT.709 ↩

  5. ITU-R BT.2020 parameter values for UHDTV. International Telecommunication Union. https://www.itu.int/rec/R-REC-BT.2020 ↩

  6. 24 fps standardised 1927–1930 by Hollywood for sound-on-film economics. Why are movies 24 frames per second?, StudioBinder, accessed May 2026. https://www.studiobinder.com/blog/why-are-movies-24-frames-per-second/ ↩

  7. Colour gamut coverage: BT.709 ≈ 35.9%, DCI-P3 ≈ 53.6%, BT.2020 ≈ 75.8% of human-visible colour. Understanding color gamut, BenQ, accessed May 2026. https://www.benq.com/en-us/business/resource/trends/understanding-color-gamut.html ↩

  8. YouTube recommended upload encoding settings: 1080p SDR 8 Mbps (12 Mbps at 48–60 fps), 4K SDR 35–45 Mbps (53–68 Mbps); H.264 High Profile, progressive, 4:2:0, MP4. YouTube recommended upload encoding settings, YouTube Help, accessed October 2026. https://support.google.com/youtube/answer/1722171 ↩

  9. Netflix internet speed recommendations: 3 Mbps HD, 5 Mbps Full HD, 15 Mbps Ultra HD 4K. Internet connection speed recommendations, Netflix Help Center, accessed October 2026. https://help.netflix.com/en/node/306 ↩

  10. Codec vs container distinction; H.264 inside MP4. Video Codec vs Container, Callaba. https://callaba.io/difference-between-video-codecs-and-containers ↩

  11. Video traffic was around 75% of all mobile data traffic at the end of 2025. Total mobile network data traffic exceeded 220 EB per month in Q2 2026, up 23% year on year. Ericsson Mobility Report, June 2026 — mobile traffic update, Ericsson, accessed October 2026. https://www.ericsson.com/en/reports-and-papers/mobility-report/dataforecasts/mobile-traffic-update ↩

  12. Sony D1 (1986), the first commercial digital video recording format, stored uncompressed SD component video. Wikipedia: Digital video, accessed October 2026. https://en.wikipedia.org/wiki/Digital_video ↩

  13. Singapore leads fixed broadband at a median 462.50 Mbps download, Ookla Speedtest Global Index for August 2026 (published 13 October 2026). Korea ranks 22nd in fixed internet speed, Seoul Economic Daily, 13 October 2026. https://en.sedaily.com/finance/2026/10/13/korea-ranks-22nd-in-fixed-internet-speed-behind-vietnam-and ↩

  14. AV1 now powers about 30% of Netflix streaming (Netflix TechBlog, December 2025). Netflix says AV1 codec now powers 30% of Netflix VOD streaming, Streaming Media Blog, 23 December 2025. https://www.streamingmediablog.com/2025/12/netflix-av1.html ↩

  15. Alliance for Open Media releases AV2 codec, press release, 9 June 2026. AOMedia, accessed October 2026. https://aomedia.org/press%20releases/Alliance-for-Open-Media-Releases-AV2-Codec/ ↩

  16. ITU-T G.114 one-way transmission time: up to 150 ms acceptable for most user applications, 400 ms upper planning limit. ITU-T Recommendation G.114, International Telecommunication Union. https://www.itu.int/rec/T-REC-G.114 ↩

  17. HDMI Forum releases Version 2.2 of the HDMI Specification: up to 96 Gbps with Ultra96 cables, June 2025. HDMI Forum, accessed November 2026. https://hdmiforum.org/hdmi-forum-releases-version-2-2-of-the-hdmi-specification ↩

  18. Chips&Media completes development of next-gen AV2 HW decoder IP (WAVE series, up to 4K240 / 8K60), announced 12 May 2026. AnySilicon, 14 May 2026. https://anysilicon.com/news/chipsmedia-completes-development-of-next-gen-av2-hw-decoder-ip/ ↩

  19. VideoLAN announces dav2d as an open-source and fast AV2 decoder, to cover the temporary lack of AV2 hardware decoding. Linux Today, 14 July 2026. https://www.linuxtoday.com/blog/videolan-announces-dav2d-as-an-open-source-and-super-fast-av2-decoder/ ↩

  20. YouTube: “AV2 streams will be available for Web as soon as early 2027”; over 90% of the catalogue (weighted by watch time) available in AV1 as of 19 March 2026 (Des Keane, YouTube). YouTube to commence AV2 streaming in 2027, TV Tech, 27 October 2026. https://www.tvtechnology.com/platform/streaming/youtube-to-commence-av2-streaming-in-2027 ↩