Client Presentation

NTT DATA
NTT DATA

HYPER2

© 2025 NTT DATA, Inc.

 What is HYPER?

HYPER Transcoding Technology is an advanced VOD transcoder that enables the creation of ultra-
optimized, high quality encoding pipelines, leveraging on innovative features such as psycho-visual pre-
filtering, hybrid-cloud architecture and Content-Aware Encoding 2.0 powered by ML algorithms.

2
© 2025 NTT DATA, Inc.
 Hyper VOD Transcoder - Overview

NTT DATA HYPER is a state-of-the-art video transcoding solution, with

an advanced batch transcoder at its core, which empowers broadcasters to
create ultra-optimized, high-quality encoding pipelines, leveraging on
innovative features such as psyco-visual pre-filtering (equalization logic to
compensate loss of details during scaling and compression), Content-
Aware Encoding powered by AI algorithms (modulates encoding
parameters according to the content complexity) and hybrid-cloud
paradigm.

HYPER optimizes bandwidth data flow, maximizing quality perception

on a specific device: with HYPER Content-Aware Encoding CAE it’s From our OTT observatory findings, as well as from independent
possible to use the bitrate strictly necessary to obtain a desired level of assessments by our clients, Hyper has the best performance in
quality on a specific screen size, eliminating the waste typical of CBR and terms of quality/bitrate, like those of Netflix's super-optimized.
VBR rate control and enables high definition even at lower bitrates. When
mixed with Advanced Player Heuristics, CAE can contribute to a robust
increase in final QoE.

2007 2012 2014 2016 2017 2019 2022 2024

[Link]: First The Customer First implementation iCE – Content-Aware 4K MLv2 profiles Hyper 2.0: complete Benchmark against
catch-up OTT chooses HYPER of Content-Aware intelligent Transcoding with ML and SynthEYE 4K refactoring, AWS Mediaconvert
service in Italy. for its first Transcoding for STB Cloud algorithms. Perceptual- dockerized, result better
OTT initiative. and OTT Devices. Encoding aware internal quality optimized for cloud quality/bitrate and
metric: SynthEYE lower cost with Hyper

3
© 2025 NTT DATA, Inc.
Video Quality of Experience (VQoE) | HYPER VOD TRANSCODER

Hyper VOD Transcoder features

Built around a sophisticated hybrid-cloud architecture (mixing both on-premise and private/public cloud transcoding), HYPER can be
configured to handle multiple queues with different priorities, levels of parallelism and encoding power. The cloud transcoding option
helps to indefinitely scale or to easily handle peak workloads.

iCE – Intelligent Cloud Encoding Other key features

The 2step paradigm (iCE – Intelligent Cloud Encoding - patented) • Optimised transcoding in H264 and H265 SDR, HDR10, HLG
accelerates the transfer of high-quality source mezzanines for • Psyco-Visual filtering for higher perceptual quality;
cloud transcoding and powers content-aware transcoding. • Segmented parallel encoding reduces encoding time of single
asset preserving excellent quality;
Content Aware Transcoding 3.0 • Flexible queues and distributed transcoding management;
• Distributed architecture for high availability;
With its Content-Aware Transcoding 3.0 powered by AI algorithms, • Caching of encoded segments and stitching algorithms allows
HYPER modulates encoding parameters, control of details and quick and cost-effective content repurposing;
video filtering according to content complexity, in order to achieve • Flexible and customizable leveraging NTT Data’s leadership
a desired perceptual quality. Encoding is therefore no more static in system integration
but dynamic and the gain in efficiency can be spent to reduce
bandwidth consumption, improve Video Quality / QoE and enable
4K at lower bitrates.

4
© 2025 NTT DATA, Inc.
Hyper VOD Transcoder - Solution

High-Level Architecture (Platform and Componentes)

Part 1/3
The distributed architecture for high availability of HYPER is designed to optimize AVC/HEVC/AV1 transcoder (up to 30% more efficient than competitors), to improve
efficiency of 20-30% applying Content-Aware Transcoding 2.0 feature, to guarantee flexibility using customizable queues and to optimize throughput and/or encoding
time. The Hybrid Cloud on-prem architecture provides high scalability and low operating costs.
HYPER is a batch transcoder with the following main domain areas: Hyper Core Services, Hyper Controllers and Encoding Profiles, Hyper Dispatcher, Hyper Encoding
Agents and Storage.

Infrastructure AWS, Google Cloud

SaaS/ Client Cloud/ On Client Cloud / On Premise

Premise

Operation Service 24x7 Yes

Single/Multi Tenant: Single Tenant

Technologies/Languages/Fra C/C++, [Link], Golang,

meworks Phyton
Users/Systems Not applicable

Setup time period 1 week

© 2025 NTT DATA, Inc.

 Example of Hybrid Cloud Deployment in a typical encoding workflow

Hyper VOD Transcoder - Solution ❑ An external user interacts with Hyper over one of its core control services
(1). A workflow manager, for example, programmatically submits new
encoding jobs using the REST API or a human operator uses the Web GUI
to interactively submit and monitor jobs. In both cases Hyper REST
High-Level Architecture (Platform and Componentes) Services is the module that will submit new jobs to the Hyper Queue
Part 2/3 Manager (2).

❑ Hyper Monitor dynamically scales instances of Hyper Controllers (4)

depending on the number of jobs to be processed (3). Controllers are
hosted on VMs/K8S pods, each instance can process a fixed number of
encoding jobs in parallel. Controllers need to access the Hyper Queue
Manager API over HTTP to fetch new jobs from the various queues (5).

❑ Hyper Controllers can offload the video processing to Encoding Agents. In

this case they use logical queues on a shared folder to share metadata and
mezzanine segments with the Dispatcher. This enables the segmented
parallel video encoding acceleration.

❑ Hyper Dispatcher monitors such logical queues and dispatch sub-jobs to

Hyper Encoding Agents specialized in segmented video encoding (7).
Hyper Monitor monitors Dispatcher demand and starts/stops Hyper
Encoding Agents as new VMs/K8S pods or terminates them if not required
anymore (4).

❑ When all job segments have been encoded, the Controller joins all
segments into a single elementary stream file and continues processing
until the final packetization files are eventually saved to a high-performance
storage or the S3 object storage.

❑ Controllers and Encoding Agents need to access the storage mainly to read
video sources and write transcoded video files. Those read and write
operations must be fast enough to achieve a high FPS encoding process
and fully exploit the CPUs of each node.
6
© 2025 NTT DATA, Inc.
 Hyper VOD Transcoder - Solution

High-Level Architecture (Platform and Componentes)

Part 3/3
❑ The network storage block represents one or more
centralized storages where shared resources (source
and transcoded assets, profiles, configuration etc.) are
stored. Encoding Agents can either access the storage
via HTTP or by mounting it as a file system. All other
modules use the latter access mode only.

❑ REST Services and Control Panel are the entry points to

Hyper v2. REST Services expose an HTTP API that
CMS external entities (3rd party) can use to integrate
with Hyper. Control Panel is a web server that runs a
web application that users can open with a browser to
submit encoding jobs and monitor the system.

7
© 2025 NTT DATA, Inc.
Hyper VOD Transcoder ❑ Superior performance in PSNR, SSIM, VMAF, IMAX SVS
and Subjective Panels assessment both in CAE and CBR
Quality Benchmark encoding scenarios.
On-going Innovations
Content-Aware Encoding 3.0
Current HYPER release offers an optional GenAI-powered psyco-visual filter that
increases furtherly VMAF (including vmaf-neg), IMAX XVS-NR and subjective
panel MOS.
GenAI can reduce artifacts and 'develop' details that in the original image are
barely hinted at or almost disappeared due to intermediate encodings or low
sensor sensitivity.
New details are perfectly integrated and increase the apparent resolution of the
video.

TV 3.0 Codecs:
Next Hyper release will include next-gen codecs such as VVC and LCEVC (over
H.264, H.265 and H.266) for vídeo and MPEG-H for audio compression.
NTT DATA HYPER
 Hyper2 Roadmap
Features and New Functionalities over FY 2025-2026 timeline towards DTV+
Use cases on Distribution, Production and Archiving across the world.

Beta New
versions release

Oct-25 Mar-26 Oct-26 Mar-27

V.2.4 V.3.0 V.3.1 V.3.2

• Dolby Atmos • Integration of LC-EVC video codec on • HDR10 on MLv3*
• MLv3 for VVC
• Audio Description h.264 & h.265 • AMIs on AWS
• LC-EVC on VVC
• VSCode editor integration in • Full VVC support marketplace*
• GCP automated deployment
GUI • Integration of MPEG-H audio codec
(Infrastructure as a code)
• Kubernates autoscaler • Cognitive services added to Hyper:
• New Analytics o Skip intro
• PID based autoscaler o Next episode
• MLv3 for AV1 o ADV insertion point
• Initial VVC support

* TV 3.0 (DTV+) requirements

10
© 2025 NTT DATA, Inc.
Video Quality of Experience (VQoE) | HYPER VOD TRANSCODER

Hyper VOD Transcoder key results

1M+ Jobs processed by Hyper

~170K hours of video (+6%)

2024 outcomes
~10M segments dispatched, encoded,
stitched

40% Overall Cache Hit (up from 28%)

52% Current Cache Hit (December 2024)

11
© 2025 NTT DATA, Inc.