What Is HDR Video? Formats, How It Works, and How to Stream It

A standard dynamic range video is mastered for a display that peaks at 100 nits. Modern HDR content is graded for 1,000 to 4,000 nits, and the format itself has headroom up to 10,000. That single number gap explains why HDR video looks like a different medium rather than a slightly better one.

HDR video carries more brightness and colour information per pixel than the SDR pipeline most streaming stacks were built around. More brightness data means visible detail in a sunset and inside a shadow at the same time, instead of one or the other.

But HDR isn’t a checkbox. It’s a chain: capture, grade, metadata, encode, package, deliver, tone map. Break any link and viewers get washed-out grey footage or, worse, a black screen.

Get the chain right and HDR buys you more visible quality per megabit than another resolution step will. Here’s how the four formats differ, which codecs carry them, and what encoding and shipping HDR actually involves.

What Is HDR Video?

HDR video is video encoded with a wider range of luminance and colour values than standard dynamic range video, using at least 10-bit colour depth and a transfer function built for high brightness. The result is brighter highlights, deeper blacks, and more visible steps between them.

Three properties define it:

• Peak luminance rises from SDR’s 100-nit reference to 1,000 nits or more
• Bit depth moves from 8 bits per channel to 10 or 12, multiplying the available shades from 16.7 million to over a billion
• Colour primaries widen from Rec.709 to DCI-P3 or Rec.2020, so deep reds and cyans stop clipping

HDR exists because SDR was designed around cathode-ray tubes. The gamma curve behind Rec.709 was reverse-engineered from CRT phosphor behaviour in the 1990s, and it tops out around 100 nits because that’s what those tubes did.

Displays now hit 1,000 to 4,000 nits, and the old container has no way to describe that light.

So the industry defined new ones. ITU-R BT.2100 is the standard that specifies HDR television, including both transfer functions in use today and the Rec.2020 colour space they ride in.

HDR vs SDR Video: What Actually Changes

The difference isn’t resolution. A 1080p HDR file and a 1080p SDR file have identical pixel counts. What changes is how much information each pixel holds and how the display is told to interpret it.

Bit depth does more work here than people expect. Stretch an 8-bit signal across a 1,000-nit range and you get banding: visible steps in gradients like skies and studio backdrops. 10-bit gives you four times as many levels to spend on that range, which is why 10-bit is a hard floor for HDR rather than a nice-to-have.

The transfer function is the other half. Gamma describes light in absolute terms that stop making sense above 100 nits. PQ, the perceptual quantizer, allocates code values according to how human vision actually discriminates brightness, so it can describe 10,000 nits in 10 bits without wasting precision on differences nobody can see.

The Four Things Every HDR Video Needs

Any HDR pipeline has to get four properties right and signal all four downstream. Miss one and playback degrades in ways that are hard to debug.

1. Peak brightness and dynamic range

Dynamic range is the distance between the darkest and brightest detail a frame can hold, measured in stops. SDR delivers roughly 6 stops. HDR formats target 14 or more, against about 20 for the human eye.

Cinema cameras have captured 14+ stops for over a decade. HDR is the first delivery format that can carry it.

2. Bit depth

10-bit is the minimum for HDR10, HDR10+, and HLG. Dolby Vision supports 12-bit. Bit depth controls quantisation, not brightness, so its job is preventing banding in the smooth gradients that HDR makes more obvious.

3. Colour primaries

Rec.2020 is the container HDR standards specify, but almost no consumer display covers it. Most grading happens inside DCI-P3, which good OLED and mini-LED panels do cover, and the P3 volume is then flagged as sitting inside a Rec.2020 container. Getting this signalling right matters more than chasing full Rec.2020 coverage.

4. Transfer function (EOTF)

Two are in use. PQ is absolute: code value 700 means a specific number of nits regardless of the display, which is why PQ content needs metadata and display-side tone mapping. HLG is relative: the signal describes a scene, and each display scales it to its own capability. That difference is exactly why HLG dominates live production and PQ dominates mastered content.

HDR Video Formats Compared: HDR10, HDR10+, Dolby Vision, and HLG

Four formats matter in production. They split along two axes: whether metadata is static or dynamic, and whether the format is royalty-free.

HDR10

HDR10 is the baseline open standard: 10-bit, PQ, Rec.2020 container, static metadata, no licensing fees. Every HDR-capable TV, console, phone, and streaming device supports it, which makes it the format you ship if you ship only one.

Its static metadata is the limitation. HDR10 sends one set of values for the whole title, carried as SMPTE ST 2086 mastering display information plus MaxCLL (maximum content light level) and MaxFALL (maximum frame-average light level). A display tone maps the entire film against that single description, so a 4,000-nit grade shown on a 600-nit panel gets one compromise curve for both a candlelit scene and a snowfield.

HDR10+

HDR10+ adds dynamic metadata to HDR10 using SMPTE ST 2094-40, sending tone mapping instructions scene by scene. The base layer stays HDR10-compatible, so a device that doesn’t understand HDR10+ falls back to plain HDR10 rather than failing.

Licensing runs through HDR10+ Technologies at nominal per-device rates. Samsung, Amazon Prime Video, and YouTube back it, but the ecosystem is smaller than Dolby’s.

Dolby Vision

Dolby Vision is the most capable and the most complicated: up to 12-bit, dynamic metadata via SMPTE ST 2094-10, and a profile system that decides backward compatibility. Profile 5 is single-layer PQ with no SDR-compatible base, Profile 8.1 uses an HDR10-compatible base layer, and Profile 8.4 uses an HLG base for broadcast.

The cost is licensing plus pipeline complexity. Dolby certifies encoders and displays, and both content and device need the licence for the dynamic metadata to be honoured. Netflix, Disney+, and Apple TV+ all ship it, and every iPhone since the 12 records it.

HLG (Hybrid Log-Gamma)

HLG was developed jointly by the BBC and NHK for live broadcast, and it carries no metadata at all. Its curve is gamma at low luminance and logarithmic in the highlights, which makes an HLG signal legible on an SDR display and correctly expanded on an HDR one.

That single-signal property is the whole point. A live production can’t grade scene by scene while the event is happening, and a broadcaster can’t run separate HDR and SDR distribution chains for a stadium feed. HLG solves both. It’s royalty-free and specified in BT.2100 alongside PQ.

If you’re picking one to start with, start with HDR10. Add HLG when you go live, and add Dolby Vision only when a distribution partner asks for it.

How Does HDR Video Work?

An HDR frame travels through seven stages, and every stage has to preserve or correctly rewrite the colour description attached to the pixels.

1. Capture. The camera records in a log or raw format holding 12 to 15 stops. Log isn’t HDR yet; it’s a wide-latitude container waiting for a grade.
2. Grade and master. A colourist works on a reference monitor of known peak brightness, usually 1,000 or 4,000 nits, and maps the log footage onto the PQ or HLG curve. The monitor’s capability becomes the mastering display metadata.
3. Measure and attach metadata. Tools scan the graded master for its brightest pixel (MaxCLL) and its brightest frame average (MaxFALL), then write those alongside the mastering display primaries. Dolby Vision and HDR10+ also generate per-scene metadata.
4. Encode. A 10-bit encoder compresses the master while writing colour primaries, transfer characteristics, and matrix coefficients into the bitstream as VUI and SEI data. Getting these three flags right is where most HDR pipelines break.
5. Package. Segments go into fragmented MP4 or CMAF containers, and the manifest declares the HDR properties so players can decide what they can handle.
6. Deliver. A CDN moves segments to viewers, with an adaptive ladder that usually carries both HDR and SDR renditions.
7. Display tone mapping. The TV or phone reads the metadata, compares it to its own peak brightness, and compresses the grade to fit. This last step is the display’s decision, not yours, which is why the same file looks different on an OLED and a laptop panel.

Steps 4 through 6 are where a streaming team lives. The rest belongs to production.

Which Codecs Support HDR Video

HDR needs a codec with a 10-bit profile and a way to carry colour signalling. That rules out most of what’s already deployed.

HEVC Main 10 is still the safe default because Apple’s ecosystem requires it and smart TVs decode it in hardware everywhere. If you already run HEVC alongside H.264, the 10-bit profile is a settings change rather than a new pipeline.

AV1 is the better long-term bet on efficiency, and VP9 Profile 2 covers Chrome and Android without licensing exposure. Most teams shipping HDR at scale run HEVC plus one royalty-free option. If codec selection is new territory, our guide to video codecs covers the tradeoffs.

Advantages of HDR Video

HDR earns its complexity in specific ways, and it helps to know which ones apply to your content.

Detail survives in highlights and shadows

This is the real win. An SDR grade forces a choice between a readable sky and a readable foreground. HDR holds both, which matters for outdoor sports, concert lighting, and anything shot against a window.

Banding drops sharply

10-bit encoding gives gradients four times the precision of 8-bit. Skies, studio walls, and slow fades stop showing the visible steps that plague 8-bit streams at moderate bitrates.

Colour volume expands

DCI-P3 covers about 46% of CIE 1931 against Rec.709’s 36%, and Rec.2020 reaches 76%. Intense stage lighting, sports kits, and vivid product colours stop clipping to flat patches.

Quality improves without more pixels

HDR at 1080p often reads as higher quality than SDR at 4K on a phone or tablet, because contrast and colour move perception harder than pixel count at normal viewing distances. That’s a useful trade when bandwidth is tight.

10-bit encoding is close to bitrate-neutral

10-bit HEVC frequently compresses as well as or better than 8-bit at matched perceptual quality, because cleaner internal precision improves prediction. Budget roughly 10–20% more bits than the equivalent SDR rendition to carry the extra highlight and shadow detail, not the 2x some teams assume.

Live HDR is genuinely practical now

HLG removed the two blockers for live: no per-scene grading and no separate SDR chain. Broadcasters have run HDR sports on a single HLG feed for years, and the same approach works for a 4K live encoder feeding an internet audience.

Disadvantages and Gotchas of HDR Video

HDR breaks in ways SDR doesn’t, and most of the failures are signalling problems rather than quality problems.

Wrong flags produce washed-out video

If the encoder writes Rec.709 primaries onto Rec.2020 pixels, or omits the transfer characteristics entirely, players interpret HDR data as SDR. The result is flat, grey, desaturated footage.

Nothing errors out, which makes this the most common and most annoying HDR bug. Fix it by asserting colour flags in your encode step and checking them in the output bitstream.

You end up shipping two ladders

Most audiences are mixed, so you carry HDR renditions for capable devices and tone-mapped SDR renditions for everyone else. That roughly doubles transcoding work and storage for the affected titles. Shared audio and a shared segment duration keep the packaging cost down.

Display tone mapping is out of your hands

You grade for 1,000 nits; a viewer watches on a 400-nit tablet. What that panel does with your highlights is its own algorithm.

Grade with a conservative MaxCLL and check the result on a low-brightness display. Your reference monitor is lying to you about the median viewer.

RTMP can’t carry it

Classic RTMP is limited to H.264 and AAC in practice, so it has no path for 10-bit HEVC or HDR metadata. Enhanced RTMP adds HEVC and AV1, but support is uneven across encoders and servers. For HDR contribution, SRT or MPEG-TS is the reliable choice today.

Editing and QA need colour-managed tools

An HDR timeline in a non-colour-managed editor produces a broken master, and you can’t judge HDR on a standard monitor. Add an HDR reference display or accept that you’re checking metadata rather than looking at pictures. Objective metrics help: VMAF has an HDR-capable variant for comparing renditions.

Device fragmentation is still real

Format support splits by platform, browser support for HEVC is inconsistent, and older HDR TVs mishandle edge cases. You’ll be maintaining a device matrix and a set of HDR test clips for regression checks.

Now that the tradeoffs are clear, here’s the practical side: what you actually run to encode, package, and deliver HDR video, and what the stack around it looks like.

How to Encode and Stream HDR Video

Six steps take an HDR master to a playing stream. The signalling steps matter more than the encoder settings.

1. Start from a correctly graded master

Your input needs to already be HDR: PQ or HLG, 10-bit, with known mastering display characteristics. You can’t create HDR by transcoding an SDR file, and inverse tone mapping produces artefacts that look worse than the original. Get the mastering display primaries and peak luminance from whoever graded it.

2. Encode with a 10-bit profile and explicit colour flags

For HDR10 in HEVC, the profile is Main 10, the pixel format is 10-bit, and the colour description has to be written into the stream:

ffmpeg -i master_pq.mov \
  -c:v libx265 -profile:v main10 -pix_fmt yuv420p10le \
  -preset slow -crf 20 \
  -x265-params "colorprim=bt2020:transfer=smpte2084:colormatrix=bt2020nc:\
master-display=G(13250,34500)B(7500,3000)R(34000,16000)WP(15635,16450)L(10000000,1):\
max-cll=1000,400:hdr10-opt=1:repeat-headers=1" \
  -c:a copy hdr10_2160p.mp4

Three parameters do the signalling work: colorprim, transfer, and colormatrix. master-display carries the ST 2086 values and max-cll carries MaxCLL and MaxFALL. For HLG, swap transfer=smpte2084 for transfer=arib-std-b67 and drop the mastering display values.

3. Generate a tone-mapped SDR ladder

Every HDR title needs an SDR fallback. Tone map from the HDR master rather than re-grading, and do the conversion in linear light:

ffmpeg -i hdr10_2160p.mp4 \
  -vf "zscale=t=linear:npl=100,format=gbrpf32le,zscale=p=bt709,\
tonemap=hable:desat=0,zscale=t=bt709:m=bt709:r=tv,format=yuv420p" \
  -c:v libx264 -preset slow -crf 20 sdr_1080p.mp4

Check faces and skies in the output. hable is a reasonable default, but heavily graded content sometimes needs mobius or a custom curve.

4. Package into fMP4 and signal HDR in the manifest

HEVC in HLS requires fragmented MP4 segments, so CMAF is the natural container and lets one set of segments serve both HLS and DASH. In HLS, the VIDEO-RANGE attribute tells players what they’re getting, and Apple’s HLS authoring specification requires it on any variant that isn’t SDR:

#EXTM3U
#EXT-X-VERSION:8

#EXT-X-STREAM-INF:BANDWIDTH=18000000,CODECS="hvc1.2.4.L153.B0,mp4a.40.2",RESOLUTION=3840x2160,FRAME-RATE=59.940,VIDEO-RANGE=PQ
hdr/2160p.m3u8

#EXT-X-STREAM-INF:BANDWIDTH=9000000,CODECS="hvc1.2.4.L150.B0,mp4a.40.2",RESOLUTION=2560x1440,FRAME-RATE=59.940,VIDEO-RANGE=PQ
hdr/1440p.m3u8

#EXT-X-STREAM-INF:BANDWIDTH=5000000,CODECS="avc1.640028,mp4a.40.2",RESOLUTION=1920x1080,FRAME-RATE=59.940,VIDEO-RANGE=SDR
sdr/1080p.m3u8

VIDEO-RANGE takes SDR, HLG, or PQ. Dolby Vision variants add a SUPPLEMENTAL-CODECS attribute. In DASH, the equivalent signalling goes in SupplementalProperty descriptors using CICP values: transfer characteristics 16 for PQ and 18 for HLG, colour primaries 9 for Rec.2020. Our guides to HLS streaming and MPEG-DASH cover the manifest structures in more depth.

5. Build a mixed HDR and SDR bitrate ladder

Keep HDR renditions at the top of the ladder and SDR below, so adaptive bitrate streaming can drop a struggling connection to a rendition that still plays. Aim for 15–25 Mbps at 2160p HDR in HEVC and 8–12 Mbps at 1440p. Our streaming bitrate reference has starting points per resolution.

Don’t mix VIDEO-RANGE values inside one rendition group without testing. Some players handle HDR-to-SDR switching mid-playback badly.

6. Run the encode, package, and delivery layer as infrastructure

Ten-bit transcoders, a packager that writes correct HDR signalling, an ABR ladder, and multi-CDN delivery take most teams six to nine months to build. None of that work differentiates your product.

That’s the layer LiveAPI handles. You get RTMP and SRT ingest, instant encoding, ABR ladders, HLS output, delivery across Akamai, Cloudflare, and Fastly, an embeddable HTML5 player, and live-to-VOD recording at up to 4K, all behind a few lines of code.

Your team keeps the parts that need your judgement: the colour pipeline, the metadata, and how far you tone map for SDR viewers. If you need specific HDR profiles in the transcode path, confirm the details with the team before you commit to a ladder design. The video transcoding API reference is the place to start.

What You Need in Your HDR Streaming Stack

Six components carry HDR from lens to screen. Each one can silently drop the colour description, so each one needs verifying.

Camera and contribution encoder

The camera needs 10-bit output and a log or HDR recording mode. The contribution encoder needs a transport that can carry 10-bit HEVC, which in practice means SRT or MPEG-TS rather than classic RTMP.

Colour-managed grading and QA

A reference monitor of known peak brightness, an editor with proper colour management, and a set of HDR test files. Without a calibrated display you’re checking metadata, not images, and that’s how washed-out masters ship.

A 10-bit transcoder

Your transcoder needs Main 10 support and the ability to pass through or rewrite colour flags. Many older transcoding setups strip SEI metadata silently, which turns a correct master into flat SDR-looking output.

A packager that writes HDR signalling

CMAF or fMP4 output with VIDEO-RANGE in HLS manifests and CICP descriptors in DASH MPDs. Check the manifest output rather than trusting the packager’s documentation.

CDN delivery sized for larger segments

HDR renditions at 4K are the largest objects in your library, so cache hit ratio and origin shielding matter more than usual. Multiple providers reduce the risk of one network becoming your ceiling during a live event. Our CDN guide covers the selection criteria.

An HDR-capable player

Web playback depends on the browser: HEVC HDR works in Safari and Edge, VP9 and AV1 HDR in Chrome and Firefox. Shaka Player and hls.js handle manifest-level HDR signalling, and native players on iOS, Android, and TV platforms generally do the right thing when the manifest is correct. If you’re building an OTT platform, plan for per-platform playback code.

Is HDR Video Worth It for Your Project?

HDR pays off for some content and adds cost for the rest. Use the content, not the trend, to decide.

Worth it if:

• Your content has real dynamic range: sports, concerts, outdoor footage, cinematic drama
• Viewers watch on HDR-capable TVs, recent phones, or premium tablets
• You already run HEVC or AV1, so 10-bit is an incremental change
• Production already grades in HDR, or can
• You compete on picture quality against services that ship HDR
• You’re launching a premium tier and need a visible differentiator

Not worth it if:

• Your content is screen recordings, slides, talking heads, or webcam footage with little range
• Most playback happens on older Android devices or corporate laptops
• Nobody in your production chain can grade or verify HDR
• Your bandwidth or storage budget is already the constraint
• You’d be shipping HDR-flagged SDR content, which looks worse than honest SDR

A middle path works well: ship SDR everywhere, add HDR to your top-tier titles or live events, and measure whether engagement moves before rebuilding the whole catalogue.

HDR Video FAQ

What is HDR in a video?

HDR in a video means the file carries brightness and colour values beyond what standard dynamic range can describe, using 10-bit or deeper colour and a PQ or HLG transfer function. On a capable display that produces brighter highlights, deeper blacks, and more visible detail at both ends. On an incapable display it needs tone mapping or an SDR version.

Is HDR video better than 4K?

They measure different things, and HDR usually has more visible effect. 4K adds pixels; HDR adds brightness range and colour. On a phone or a TV at typical viewing distance, 1080p HDR often looks better than 4K SDR, because contrast moves perception more than resolution once pixels are small enough.

Should I turn HDR video on or off when recording?

Leave it on if your editing and delivery chain handles HDR end to end. Turn it off if you’re shooting quick clips for social platforms or editing in a tool without colour management, because a mishandled HDR file plays back grey and washed out on other people’s devices.

Does HDR mean better quality?

Only when the whole chain is correct. Properly graded HDR with matching metadata looks clearly better than SDR. HDR-flagged content that was upconverted from SDR, or HDR with wrong colour flags, looks worse than the SDR original. The label isn’t the quality.

Does YouTube support HDR?

Yes. YouTube accepts HDR10 and HLG uploads and transcodes them to VP9 Profile 2 and AV1 for playback. Its HDR upload requirements ask for 10-bit video with correct Rec.2020 PQ or HLG signalling, and files without that signalling get treated as SDR.

How do I remove HDR from a video?

Tone map it to SDR rather than just stripping the metadata. Removing the flags alone leaves PQ-encoded pixels being read as gamma, which produces flat grey footage. The ffmpeg zscale and tonemap chain in the encoding section above converts properly in linear light.

What HDR format does iPhone record?

iPhone 12 and later record Dolby Vision Profile 8.4, which is 10-bit HEVC with an HLG-compatible base layer. That means the footage plays correctly on non-Dolby displays as HLG, which makes it unusually friendly to general-purpose pipelines compared with Profile 5.

Do I need 4K for HDR video?

No. HDR and resolution are independent, and 1080p HDR is a valid and useful combination. It’s a good option when bandwidth is limited, since you get most of the perceptual benefit at a fraction of the bitrate of 4K HDR.

How much extra bandwidth does HDR video need?

Plan for roughly 10–20% more than the equivalent SDR rendition at the same resolution. 10-bit HEVC compresses about as efficiently as 8-bit, so the extra bits pay for real highlight and shadow detail rather than for the deeper container itself.

Can I stream HDR video live?

Yes, and HLG is how. It needs no per-scene metadata and stays legible on SDR displays, so a single feed serves a mixed audience. The contribution path has to carry 10-bit HEVC, which means SRT or MPEG-TS rather than classic RTMP.

Closing Thoughts on HDR Video

HDR video is a bigger visible upgrade than another resolution step, and the reason it’s harder is that brightness and colour have to be described accurately at every stage rather than just carried. Most HDR failures in production aren’t quality problems. They’re a missing transfer characteristic flag or a transcoder that dropped SEI metadata.

So treat signalling as the deliverable. Verify colour primaries, transfer function, and matrix coefficients in your encoded bitstream and your manifests, ship a tone-mapped SDR ladder alongside HDR, and test on a dim display as well as a reference monitor.

Ready to ship HDR-capable video without building the pipeline? LiveAPI handles RTMP and SRT ingest, instant encoding, ABR ladders, HLS delivery over Akamai, Cloudflare, and Fastly, and live-to-VOD at up to 4K, so your team spends its time on the colour pipeline instead of the plumbing. Get started with LiveAPI.