Per-Shot Encoding

IRIS Adaptive Per-Shot Encoding analyzes every shot individually and allocates bits exactly where the picture needs them. Combined with IRIS Denoising, which removes noise before the encoder ever sees the signal, this delivers higher VMAF, fewer quality drops in demanding scenes and a more consistent viewing experience across the entire stream.

Industry reference: Netflix reports around 30% bitrate savings for its shot-based Dynamic Optimizer. IRIS per-shot encoding targets the same range, up to 30%, in a single pass without iterative encoding loops.

UHD in HEVC at 7.54 Mbit/s with 96.02 VMAF

April 29, 2026

Test material: Netflix "Meridian", encoded in HEVC and delivered via HLS in UHD at 59.94 fps.

7.54 Mbit/sBitrate, UHD @ 59.94 fps
96.02VMAF
5.06QBVE, lower is better
HEVC / HLSCodec and delivery

How? By removing noise from the signal before the encoder ever sees it, combined with IRIS Adaptive Per-Shot Encoding, which analyzes every shot individually and allocates bits with surgical precision. The result: higher VMAF, fewer quality drops in demanding scenes and a more consistent viewing experience across the entire stream.

Our customers see this every day, on their own material. If you want to see what it does to yours, let's talk.

#Streaming #HEVC #HLS #UHD #Broadcast #OTT #PerShotEncoding #VideoQuality #IRIS #VMAF #QBVE #WaveletBeam

How IRIS Per-Shot Encoding Works

  1. Denoise first: IRIS Denoising cleans the signal before it hits the encoder, so bits are spent on detail, not on noise.
  2. Analyze every shot: IRIS.ANALYST generates precise metadata for each shot in a single pass, without iterative VMAF loops.
  3. Allocate bits per shot: Every scene gets exactly the bitrate its complexity requires, for up to 30% bitrate savings.
  4. Deliver via HLS and DASH: Dynamic per-shot encoding in HEVC or AV1. AV1 with our in-house Film Grain Synthesis is already integrated, with film grain tables we developed from scratch to avoid repeating patterns and to match the original grain.

The entire pipeline runs as a fully automated process with flexible Kubernetes integration, as well as via Docker and VMs.

Test it on your material: info@waveletbeam.com Per Shot Encoding explained AI

Per-Shot vs. Per-Title: Why One-Path is the Future

January 31, 2026

Traditional Per-Title encoding relies on exhaustive VMAF-driven loops that consume massive computing power and time. At Wavelet Beam, we believe efficiency should be inherent, not iterative. Our One-Path philosophy eliminates redundant analysis cycles.

The Wavelet Beam Advantage: Intelligent Precision

By leveraging IRIS.ANALYST, we generate precise metadata in a single step. This allows our Per-Shot Encoding to adapt instantly to the unique complexity of every scene, ensuring optimal quality without the overhead of multiple encoding passes.

  • IRIS-Denoising Integration: We clean the signal before it hits the encoder, ensuring bits are spent on detail, not noise.
  • One-Path Workflow: Metadata generation and encoding happen in a seamless, high-speed pipeline.
  • Up to 30% Bitrate Savings: Superior visual fidelity at significantly lower bandwidth through shot-based granularity.
#WaveletBeam #PerShotEncoding #IRISDenoising #OnePath #GreenStreaming

QBVE: Is it just a single glitch or a nightmare?

February 5, 2026

In high-end video encoding, averages can be misleading. At Wavelet Beam, we've seen cases where a video achieves a "Mean VMAF" of 95, yet contains visible quality collapses that ruin the viewer experience.

QBVE Formula

The Wavelet Beam Solution: QBVE

We utilize QBVE (Quadratic Bad VMAF Energy). Unlike standard means, QBVE functions as a Gated MSE, measuring the failure energy only when it falls below our quality floor:

QBVE Diagram Comparison

Comparison: Standard Encode (QBVE 9.46) vs. Wavelet Beam IRIS (QBVE 5.06).
Double-click to open. Single-click to zoom.

While the standard encoder (left) struggled, our IRIS per-shot optimization (right) preserved structural integrity and nearly halved the "Bad Energy."

#WaveletBeam #VMAF #QBVE #VideoQuality #IRIS

Results and Insights

Measurements and findings from our work, summarized from Dirk Hildebrandt's posts and articles.

  1. QBVE: what averages hide

    QBVE works as a gated MSE: frames above VMAF 93 contribute nothing, deviations below the gate are squared, so visible artifacts weigh far more than small fluctuations. IRIS-Denoising as a pre-process stabilizes the source, so fewer frames fall below the gate in the first place.

    Read on LinkedIn

  2. Less compute than convex hull encoding

    Convex hull encoding relies on brute force: many trial encodes to find the right parameters. A shot-based workflow with QBVE as a VMAF-based control instance reaches the target with less compute, less time and less energy.

    Read on LinkedIn

  3. When 1 VMAF point means nothing and everything

    A 10 minute video at 50 fps has 30,000 frames. If about 1,364 frames, roughly 27 seconds, collapse from VMAF 94 to 72, the VMAF mean drops by just 1 point. QBVE with a gate at VMAF 93 reports 20.05 for the same case. In addition, no systematic error intervals have been published for VMAF, so small deltas between encoders are hard to judge.

    Read on LinkedIn

  4. 30% bitrate reduction in a single pass

    Measured on real content: noise and grain alone account for up to 30% of the bitrate budget. Removing them per shot before encoding frees the encoder for real image content and enables up to 30% lower bitrates at the same QBVE quality score. No convex hull, no encoding loops, no brute force.

    Read on LinkedIn

  5. All UHD streams below 10 Mbit/s

    Per-shot adaptive encoding with IRIS.ANALYST keeps UHD streams below 10 Mbit/s without over or under encoding single scenes. Massive parallel chunk processing uses all available CPU and GPU resources, with output as HLS or DASH in H.264, HEVC or AV1.

    Read on LinkedIn

  6. UHD HEVC below 10 Mbit/s: counting bad frames

    Critical low-light scene of Netflix Meridian, both encodes at 9.6 Mbit/s, one of them with IRIS and IRIS.ANALYST (2 pass encoding at that time). Frames below VMAF 93 fell from 25.73% to 13.29%, below VMAF 87 from 3.76% to 1.94%, and below 40 dB PSNR from 51.62% to 10.15%.

    Read on LinkedIn

Frequently Asked Questions

What is per-shot encoding?

Instead of one bitrate ladder for a whole title, every shot is analyzed and encoded with exactly the bits its complexity needs. IRIS.ANALYST generates the metadata in a single pass, without iterative VMAF encoding loops.

How much bitrate can IRIS per-shot encoding save?

Up to 30%. One example: Netflix "Meridian" in UHD at 59.94 fps, encoded in HEVC and delivered via HLS at 7.54 Mbit/s with a VMAF of 96.02 and a QBVE of 5.06.

What is QBVE?

QBVE (Quadratic Bad VMAF Energy) only measures the failure energy below a quality floor, for example VMAF 93, and squares the deviation. Short, visible quality drops can therefore not hide behind a high average VMAF.

Which codecs and delivery formats are supported?

HEVC and AV1, exported for HLS and DASH. AV1 with our in-house Film Grain Synthesis is already integrated into the per-shot pipeline, which runs fully automated with Kubernetes, Docker or VMs.