10-bit HEVC from 8-bit source — when it helps and when it's pointless

10-bit HEVC encoding from 8-bit source is one of the more counterintuitive video pipeline decisions. The source has only 256 values per channel; the encoded output has 1024 values per channel — but the source data is still 8-bit. So is 10-bit encoding worth the bandwidth cost? The honest answer: yes, surprisingly, for premium SDR content. The encoder's internal precision benefits even when source is 8-bit. This page is the engineering reference.

#What 10-bit encoding from 8-bit source does

The mechanics:

• Source: 8-bit YUV (256 values per channel).
• Encoding: 10-bit YUV pixel format internally (1024 values per channel).
• Output: 10-bit HEVC stream (Main 10 profile).

The encoding doesn't add information that wasn't in the source — you can't manufacture quantization values that didn't exist. But the encoder's internal math runs at higher precision, and the output preserves more of the encoder's quality decisions.

#Why 10-bit helps from 8-bit source

The benefit comes from encoder math precision:

1. Internal calculations.

Codec encoding involves many internal operations:

• DCT and inverse DCT.
• Quantization.
• Filtering.
• Motion estimation.

Each operation in 8-bit precision accumulates rounding errors. In 10-bit precision, the rounding errors are smaller; quality is preserved better through the pipeline.

2. Smoother quantization in flat regions.

In flat regions (skies, smoke, gradients), 8-bit quantization can produce visible banding. 10-bit's finer quantization values let the encoder preserve smoother transitions even from 8-bit source.

This is the most-visible benefit. 8-bit-source content with significant flat regions encoded as 10-bit looks meaningfully smoother than encoded as 8-bit.

3. Better deblocking and filtering.

In-loop filters operate at higher precision; output is cleaner.

4. Reduced banding from compression artifacts.

8-bit encoding can introduce banding through compression artifacts. 10-bit encoding has finer values to express compression artifacts, reducing visible banding.

#Quality benefit measurement

For typical content with 8-bit source encoded at HEVC Main 10 vs Main:

Content type	VMAF improvement
Talking heads / drama	0.5-1.5 VMAF
Animation / clean 2D	1.0-2.0 VMAF
Sports (busy)	0.0-0.5 VMAF
Nature / skies	1.5-3.0 VMAF

The pattern: content with significant flat regions benefits most. Sports and other high-detail content benefits less. Premium animation and content with dynamic-range gradients benefit most.

For premium SDR streaming, the VMAF improvement is real and worth the bandwidth cost.

#Bandwidth implications

10-bit content needs ~10-20% more bitrate than equivalent-quality 8-bit content. The exact number depends on:

• Content type (flat content sees less overhead; complex content sees more).
• Encoder configuration (preset matters).
• Quality target (higher targets see more overhead).

For a streaming service shipping 10-bit HEVC at premium tiers and 8-bit HEVC at lower tiers:

• Premium tier (4K, 1080p): 10-bit, 10-20% more bitrate vs equivalent 8-bit.
• Mid tier (720p): 8-bit (no 10-bit benefit; bandwidth cost).
• Floor tier (360p, 240p): 8-bit (no 10-bit benefit; smaller resolution masks any quality difference).

The premium tier 10-bit benefits the audience that watches on capable HDR-capable displays where banding is most visible. The lower tiers stay 8-bit for bandwidth efficiency.

#Hardware decoder support

10-bit HEVC (Main 10 profile) hardware decode support:

• Smart TVs — every HEVC-capable TV from 2017+ supports Main 10. Required for HDR10 (which mandates 10-bit). Universal for HDR-capable devices.
• Mobile — every HDR-capable smartphone supports Main 10 in hardware.
• Desktop GPUs — every HEVC-decoding GPU supports Main 10.
• Browsers — Safari (macOS, iOS) supports Main 10; Chromium increasingly supports it; Firefox supports it via OS-level decoders.
• Set-top boxes — Apple TV, Roku, Chromecast support Main 10.

Effectively universal in 2026 for HDR-capable devices. The audience that needs 8-bit-only HEVC playback (no Main 10 support) is primarily older smart TVs and downlevel set-top boxes — typically small percentage of streaming audience.

For pipelines, ship Main 10 for HDR (mandatory) and premium SDR (optional but recommended); ship Main (8-bit) for floor tier compatibility with downlevel devices.

#ffmpeg 10-bit encoding

The configuration:

ffmpeg -i input_8bit.mp4 \
  -c:v libx265 -profile:v main10 -pix_fmt yuv420p10le \
  -preset slow -crf 24 \
  -c:a copy output_10bit.mp4

Components:

• -profile:v main10 — declares Main 10 profile (10-bit).
• -pix_fmt yuv420p10le — 10-bit YUV 4:2:0 little-endian. The "le" is endianness; little-endian is standard.
• -preset slow — slower preset for premium quality (more compute; better encoding).
• -crf 24 — slightly higher CRF than 8-bit because 10-bit's compression efficiency is slightly better at the same target VMAF.

For SVT-AV1 10-bit encoding from 8-bit source:

ffmpeg -i input_8bit.mp4 \
  -c:v libsvtav1 -pix_fmt yuv420p10le \
  -preset 6 -crf 28 \
  output_10bit.mp4

Same logic; the -pix_fmt yuv420p10le triggers 10-bit encoding.

#When 10-bit is mandatory

10-bit is required (not optional) for:

• HDR10 — mandates 10-bit Main 10 profile.
• Dolby Vision — typically 10-bit (Profile 8.1 baseline); 12-bit for some profiles.
• HDR10+ — mandates 10-bit.
• HLG — mandates 10-bit for compliant signaling.
• ATSC 3.0 broadcast — typically 10-bit for HDR delivery.

For HDR pipelines, 10-bit is structural. There's no "choice" between 8-bit and 10-bit for HDR; HDR requires 10-bit.

#When 10-bit isn't worth it

The cases where 10-bit's bandwidth cost isn't justified:

• Floor-tier streaming — 360p / 480p where resolution dominates quality perception; 10-bit's banding-reduction benefit is masked by resolution-related artifacts.
• Mobile-only streaming — small displays mask banding; 10-bit benefit is less visible.
• Bandwidth-constrained delivery — when every Mbps matters for the audience's connection, 8-bit's lower bitrate wins.
• Sports / high-detail content — flat-region banding isn't a concern; 10-bit benefit is small.

For these cases, 8-bit HEVC is fine. For premium streaming with mixed content types, 10-bit on top tiers and 8-bit on lower tiers is the typical mix.

#12-bit considerations

Beyond 10-bit, 12-bit content exists:

• HEVC Main 12 profile — 12-bit support.
• AV1 High profile — 12-bit support.
• VVC — 12-bit standard.

Used for:

• Dolby Vision Profile 7 (Blu-ray premium) — 12-bit.
• Cinema mastering — 12-bit common for D-Cinema.
• High-end mastering pipelines — sometimes 16-bit through working spaces, output at 12-bit.

For consumer streaming in 2026, 12-bit is rare. 10-bit is the standard for premium HDR; 12-bit is for specific Dolby Vision workflows and broadcast/cinema mastering.

#Operational considerations

Things that matter for 10-bit encoding from 8-bit source in production:

• Pixel format must be specified — -pix_fmt yuv420p10le is the trigger; without it, encoder defaults to 8-bit even with -profile:v main10.
• Source format compatibility — verify ffmpeg's source decoder produces 8-bit data; confirm conversion to 10-bit pixel format works.
• Bandwidth calibration — 10-bit ladders need their own bitrate calibration vs 8-bit. Don't reuse 8-bit calibration values.
• Quality verification — VMAF or subjective testing on representative content to confirm the 10-bit benefit is real for your content.
• Hardware decoder verification — test on actual target devices.

#Encoder-specific behavior

Different encoders handle 10-bit-from-8-bit differently:

x265:

• Most production-tested 10-bit encoder.
• Quality benefit on 8-bit source is real and well-documented.
• Compute cost vs Main is roughly +20-30%.
• Default for most production HEVC pipelines.

SVT-AV1:

• 10-bit support solid in current versions.
• Quality benefit similar to x265.
• Compute cost vs 8-bit is +10-25%.
• Common for AV1 pipelines.

libaom-av1:

• 10-bit support in cpu-used 0-4 range.
• Quality benefit highest at slow presets.
• Compute cost significant.

Hardware encoders (NVENC, Quick Sync):

• Most modern hardware encoders support 10-bit.
• Quality benefit smaller than software encoders.
• Compute cost minimal vs 8-bit hardware encode.

For production at premium quality with software encoders, 10-bit is the answer. For hardware encoding, the benefit is smaller; sometimes 8-bit hardware suffices.

#What MpegFlow does with 10-bit encoding

MpegFlow's FfmpegExecutor worker image runs HEVC Main 10 (x265) and AV1 10-bit (SVT-AV1) for 10-bit output. The DAG runtime expresses each ladder rung's bit depth as a per-stage parameter; the partitioner persists each rendition stage to job_stages with explicit dependency tracking and per-stage retry. Default templates run 10-bit for top/middle tiers and 8-bit for floor tiers.

For HDR workflows, 10-bit is mandatory; the workflow YAML enforces it for HDR encoding stages.

For SDR pipelines where customers want premium-tier 10-bit from 8-bit source, the per-rung bit-depth parameter lets the same workflow run 10-bit on premium tiers and 8-bit on floor tiers as parallel sibling stages — the partitioner emits both rungs from the same upstream source.

The strict-broker security model handles bit-depth work like any pipeline payload — workers carry no ambient credentials; content access flows through short-lived presigned URLs scoped per stage; access is disposed on completion.

For multi-codec ladders combining AV1 + HEVC + H.264, the bit-depth choices typically align: AV1 + HEVC at 10-bit for top tiers, H.264 at 8-bit for floor tiers (H.264 hardware decoders rarely support 10-bit High profile).

The general guidance: 10-bit from 8-bit source is a real benefit for premium SDR content with significant flat regions. For HDR, 10-bit is mandatory. For floor-tier streaming, 8-bit is acceptable. Match bit depth to tier and content type rather than applying universal defaults. The VMAF improvement is small but real; the bandwidth cost is also real but worth it for premium content where banding visibility matters.