Robot explorer standing in a glowing cave filled with giant bioluminescent mushrooms and molten orange formations.
Neural Technology for Mobile

Deliver cinematic visual experiences, higher frame rates,
and faster development workflows on mobile with Arm Neural Technology.

Get Started
Entitlements could not be checked due to an error reaching the service. Showing non-confidential search results only.
ON THIS PAGE
Overview

Build Desktop-Class Immersive Experiences on Mobile

Arm Neural Technology helps developers improve image quality, increase frame rates, and reduce GPU workload using AI-powered neural upscaling, denoising, and frame generation techniques.




Arm Neural Technology consists of:

  • ✦ Neural Super Sampling (NSS)
    Improves image quality and graphics efficiency.
  • ✦ Neural Frame Rate Upscaling (NFRU)
    Increases frame rate.
  • ✦ Neural Super Sampling and Denoising (NSSD)
    Enables advanced lighting and ray tracing workflows.

Deliver Cinematic Experiences

Enable advanced lighting, ray tracing, and higher visual fidelity on mobile.

extend-and-improve-icon

Improve Power Efficiency

Reduce GPU workload and free up performance budget for more ambitious content.

iterate-faster-icon

Iterate
Faster

Reduce lengthy bake cycles and accelerate lighting and content workflows.

The Future of Mobile Gaming: Neural Dawn

Neural Dawn is a production-ready game co-built by Arm and Sumo Digital on Unreal Engine 5.6.1. The world's first game to use Unreal Engine MegaLights on mobile, demonstrating what's possible on Arm Neural Technology.

Explore Neural Dawn
Start Building

Neural Graphics Development Kit

Start building with the Neural Graphics Development Kit that includes Unreal Engine plugins, tools, documentation, sample code, and model resources for resolution and frame rate upscaling.

Dev Icon Quick Step
Unreal Engine
Plugin

Add NSS to an Unreal Engine project. Render at a lower resolution, upscale with neural graphics, and compare image quality and performance in a real engine workflow.

Try NSS in Unreal
Orange Sharper Graphics icon
Neural Graphics
Models

Explore open and trainable neural graphics models for real-time resolution and frame rate upscaling workflows.

Explore Models on Hugging Face
tool icon
Vulkan
ML SDK

Integrate neural graphics into custom engines with Vulkan ML extensions for inference, optimization, emulation, and debugging.

Build with Vulkan ML
Orange Portability icon
Code
Samples 

Use sample code to understand model integration, inference setup, graph construction, and rendering pipeline integration.

Explore Sample Code
Dev Icon Framework
Neural Graphics
Model Gym

Explore, train, and evaluate neural graphics models in a guided workflow.

Try the Model Gym

Join the Early Access Program for access to Arm Neural Technology ahead of release and future model updates. For public NFRU resources, start with the NFRU Learning Path, SDK/plugin documentation, and model resources.

 
Join Early Access Program
Development Resources

Choose Your Implementation Path

  • Try in UNREAL
  • Build with Vulkan ML
  • Prepare and Tune Models

Explore NSS and NFRU model resources, fine-tune neural graphics models with your own content, and prepare models for Vulkan-compatible neural graphics workflows.

model

NSS Model on Hugging Face

Download and explore the NSS model and dataset for neural upscaling workflows.

tool

Prepare Models for Neural Graphics

Understand and debug the model preparation flow used by Arm neural technology in neural graphics pipelines.

Learning Paths icon

Fine-Tuning Models with Model Gym

Configure and train neural graphics models in Jupyter Notebooks.

documentation

Quantize Neural Upscaling Models with ExecuTorch

Reduce latency and memory bandwidth by exporting INT8 models to .vgf using the ExecuTorch Arm backend.

blog

How NSS Works

Learn how NSS works, from training and network architecture to post-processing and inference.

Documentation

Neural Super Sampling Fine-Tuning Guide

Fine-tune the NSS network with your own game content using Model Gym.

documentation

Arm Neural Upscaler Use Case Guide

Learn how NSS fits into a rendering pipeline and where neural upscaling can improve performance.

documentation

NSS in Unreal White Paper

Compare Arm ASR, FSR2, DLSS, and NSS, and learn where neural upscaling fits in mobile graphics workflows.

tools

NFRU Model on Hugging Face

Download and explore the NFRU model, checkpoints, VGF file, scenario, and license.

tools

NFRU Fine-Tuning with Model Gym

Fine-tune NFRU models using Model Gym workflows.

tools

NFRU Video

Learn how mobile Neural Frame Rate Upscaling works and how to evaluate it in neural graphics workflows.

tools

NFRU Case Study: Project Moku

Explore a corridor scene case study with NFRU benchmarks, image quality, performance, and getting started guidance.

tools

NSS Data Capture Plugin

Capture high-quality neural graphics datasets directly from Unreal Engine. Export gameplay frames and metadata to train, fine-tune, and evaluate neural graphics models using your game content.

tools

Prepare Models for Neural Technology

Prepare neural graphics models for deployment with Arm Neural Technology. Convert, optimize, and validate models for Vulkan-based inference to move seamlessly from training to real-time graphics integration.

FAQs

Frequently Asked Questions

What is Neural Technology?

Neural Technology uses machine learning to improve rendering workflows. For mobile games, this includes techniques such as Neural Super Sampling (NSS), Neural Frame Rate Upscaling (NFRU), and Neural Super Sampling and Denoising (NSSD).

What is Neural Super Sampling?

Neural Super Sampling, or NSS, is Arm’s neural upscaling technology for improving image quality and performance in mobile graphics workflows. NSS renders fewer pixels to reduce GPU workload, then uses a trained model to reconstruct a high-quality visual output. NSS resources are available through the Neural Graphics Development Kit.

What is Neural Frame Rate Upscaling?

Neural Frame Rate Upscaling (NFRU) is an Arm neural graphics technology for increasing frame rate by generating intermediate frames. Developers can evaluate NFRU through the Neural Graphics Development Kit, Unreal Engine resources, model resources, and supporting Learning Paths.

How do NSS and NFRU differ?

NSS improves image quality by reconstructing high-quality output from fewer rendered pixels. NFRU increases perceived frame rate by generating intermediate frames using neural graphics techniques. Developers can start with NSS for upscaling and then evaluate NFRU for frame rate upscaling.

What is Neural Super Sampling and Denoising (NSSD)?

Neural Denoising removes noise generated by ray tracing while preserving detail. It works together with neural upscaling to enable real-time path tracing on mobile with fewer rays per pixel.

What is Neural Dawn?

Neural Dawn is a production-quality Unreal Engine 5.6.1 mobile game created by Arm and Sumo Digital. It shows Arm Neural Technology running in a real mobile graphics workflow with neural upscaling, denoising, frame-rate upscaling, MegaLights, and mobile ray tracing.

Do I need a custom engine to try NSS?

No. Unreal developers can start with the relevant Unreal Engine plugin and Learning Path. Use the NSS plugin for neural upscaling workflows, or the NFRU plugin for frame rate upscaling workflows. Developers who need more control can move into Vulkan ML tools and custom engine integration.

Where should I start?

Start with the technology that matches your goal. Use Neural Super Sampling (NSS) in Unreal Engine to improve image quality while maintaining performance. Use Neural Frame Rate Upscaling (NFRU) when your priority is smoother motion and higher frame rates. For custom engines, start with the Vulkan ML tools and sample code. To train or optimize your own models, explore the model preparation resources and Model Training Gym Learning Paths.

What is Vulkan ML used for?

Vulkan ML helps developers integrate neural technology into custom rendering pipelines, including inference, graph setup, model conversion, emulation, and debugging.

Can I train or fine-tune models?

Yes. Use the model resources, Prepare models for neural graphics with Arm Neural Technology Learning Path, and Model Gym workflows to explore model architecture, preparation, fine-tuning, and content-specific optimization.

Do I still need to join the Early Access Program to try NFRU?

No, you don’t have to join the Early Access Program, but you can join if you want access to pre-release neural technologies and future model updates. For public NFRU resources, start with the NFRU Learning Path, SDK/plugin documentation, and model resources.

Copyright © 1995-2026 Arm Limited (or its affiliates). All rights reserved.