Video Transcoding &
Streaming Pipeline Engineering
From ingest to playback: adaptive streaming pipelines built and QC'd against the published Apple HLS Authoring, Netflix Photon and Amazon Prime Video delivery specifications. Hardware encoding, multi-DRM and SCTE-35 ad insertion built for production scale.
What we engineer
Every capability below is in production use, described in the terms the specifications themselves use.
Adaptive streaming pipelines
HLS, LL-HLS, CMAF, DASH end-to-end. Per-title encoding ladders, ABR logic, and manifest generation built and QC'd against the published Apple HLS Authoring, Netflix Photon and Amazon Prime Video delivery specifications.
Hardware & software encoders
NETINT Quadra VPU T1A/T2A, NVENC, Intel Quick Sync, AMD VCN for hardware acceleration. x264, x265, SVT-AV1, VVenC for software paths. Codec selection matched to latency and cost targets.
Multi-DRM content protection
FairPlay, Widevine and PlayReady with CENC and cbcs encryption schemes. Key rotation, licence server integration and PSSH box generation.
Broadcast & ad insertion
IMF (SMPTE ST 2067) packaging, HLS Interstitials, SCTE-35/104 cue injection and AI-driven ad-break detection on content signals.
Quality metrics & QC
VMAF, PSNR, SSIM computed per segment and surfaced in an operations dashboard. Automated QC gates before segment packaging.
Observability
OpenTelemetry tracing through encode → package → origin. Kafka event streams for segment-level telemetry. Alerts on stall rate, latency and error spikes.
Transcoding pipeline dashboard — live demo
Interactive simulation of a video transcoding farm: 8 worker nodes running ABR ladder generation, live job progress, worker throughput and quality gates. The pipeline advances every 2 seconds.
GB/h processed across all encode and packaging workers.
- Current
- 96 GB/h 3 active · 7 queued
- Peak today
- 120 GB/h Ø 87 GB/h
- Completed today
- 41 1 failed · 97.4% success
Live progress
Queue and in-flight jobs. Progress advances while the simulation is running.
feature_reel_master.mov
job-2401 · HLS 1080p H.264 · worker-01
promo_cutdown_v3.mp4
job-2402 · HLS 720p HEVC · worker-02
trailer_uhd_source.mxf
job-2403 · MP4 4K HEVC · worker-04
episode_s02e04.mp4
job-2404 · HLS 1080p H.264
sizzle_alt_audio.mov
job-2405 · AAC 5.1
Pipelines
Saved pipelines. Run starts a local mock job when a worker is idle.
/ingest/features → HLS
/ingest/promos → CMAF
/ingest/mezzanine → MP4
Workers
Connected encode and packaging workers. Busy nodes track the assigned job progress.
worker-01
10.0.12.11 · transcode
HLS 1080p H.264
worker-02
10.0.12.12 · transcode
HLS 720p HEVC
worker-03
10.0.12.13 · transcode
Waiting for the next job
worker-04
10.0.12.14 · packaging
MP4 4K HEVC
Video profiles
Encode ladders used by pipelines and manual uploads.
| Name | Resolution | Codec | Bitrate | Frame rate |
|---|---|---|---|---|
| HLS 1080p H.264 | 1920×1080 | H.264 | 5 Mbps | 25 fps |
| HLS 720p HEVC | 1280×720 | H.265 | 3 Mbps | 25 fps |
| MP4 4K HEVC | 3840×2160 | H.265 | 12 Mbps | 25 fps |
| HLS 480p H.264 | 854×480 | H.264 | 1.5 Mbps | 25 fps |
Audio profiles
Audio encode presets for stereo and surround delivery.
| Name | Codec | Channels | Sample rate | Bitrate |
|---|---|---|---|---|
| AAC Stereo | AAC | 2.0 | 48 kHz | 192 kbps |
| AAC 5.1 | AAC | 5.1 | 48 kHz | 384 kbps |
| AC-3 Broadcast | AC-3 | 5.1 | 48 kHz | 448 kbps |
ABR ladder
Adaptive bitrate renditions ordered by resolution. Bar length is relative to the top rendition's bitrate.
Quality metrics (VMAF)
Per-rendition quality probes against the mezzanine. VMAF ≥ 90 passes, 80–90 watched, below 80 fails QA.
| Rendition | Resolution | Codec | Bitrate | VMAF | PSNR | SSIM | QA |
|---|---|---|---|---|---|---|---|
| 4K HEVC | 3840×2160 | H.265 | 12 Mbps | 96.8 | 44.1 | 0.987 | Pass |
| 1080p H.264 | 1920×1080 | H.264 | 5 Mbps | 94.2 | 42.6 | 0.979 | Pass |
| 1080p HEVC | 1920×1080 | H.265 | 4 Mbps | 95.1 | 43.0 | 0.982 | Pass |
| 720p HEVC | 1280×720 | H.265 | 3 Mbps | 92.4 | 40.8 | 0.969 | Pass |
| 480p H.264 | 854×480 | H.264 | 1.5 Mbps | 87.6 | 37.2 | 0.941 | Watch |
| 360p H.264 | 640×360 | H.264 | 0.8 Mbps | 81.3 | 34.1 | 0.912 | Watch |
S3 connectors
Public connector metadata only — name, region and bucket. No access keys or secrets.
eu-central-1 · media-ingest-prod
Credentials are never shown in this demo.
eu-west-1 · delivery-packaging
Credentials are never shown in this demo.
us-east-1 · media-archive-cold
Credentials are never shown in this demo.
DRM settings
Neutral packaging and key-rotation options for protected outputs.
Enabled · rotation every 24h
Provider
Widevine + FairPlay
Packaging
CMAF + HLS
Event log
Live pipeline events from scheduler, workers, packager and QA probes.
Demonstration UI composed from shacdn primitives. Not connected to a live cluster.
Formats, encoders & standards
Formats and hardware we have integrated in production.
Built against published delivery specifications
Output packages run through Netflix Photon conformance tooling. CMAF segment structure, timing metadata and codec constraints are validated against the published Photon requirements before delivery.
Manifests and segments validated against the Apple HLS Authoring Specification for HEVC, FairPlay encryption and HLS Interstitials. Bitrate ladder and timing requirements checked against the published authoring guidelines.
Encoding ladders and packaging aligned with the Amazon Prime Video documented ingest profile, including HDR metadata passthrough and audio normalisation per the published requirements.
What you get
Encoding ladder document
Per-title analysis with bitrate and resolution targets per codec, backed by measured VMAF scores and a documented rationale.
Infrastructure & hardware selection
Measured cost per hour of processed video across NETINT, NVENC and software paths, with a hardware sizing recommendation for your throughput target.
Packaging & DRM configuration
HLS/DASH/CMAF manifests, CENC/cbcs encryption, key rotation schedule and licence server integration tested end-to-end.
Operations dashboard
Live QoE and quality metrics — VMAF, stall rate, latency — with alert thresholds configured for your SLOs and surfaced in Grafana or equivalent.
Runbooks & alerting
Documented runbooks for common failure modes and on-call alert routing configured for your incident management tool.
Load-tested handover
Pipeline signed off at target throughput with a documented handover package and a one-week post-delivery support window.
Optimising a transcoding pipeline?
Share your ingest format, output targets and scale requirements — we will outline an architecture and hardware selection matched to your cost and latency goals.