AV2 vs. AV1: 2025 4K-Streaming Benchmarks Using Netflix Open Content & SimaBit Pre-processing

Introduction

The video streaming landscape is experiencing unprecedented growth, with the global media streaming market projected to reach $285.4 billion by 2034, growing at a CAGR of 10.6% from 2024's $104.2 billion. (Sima Labs) As video traffic is expected to comprise 82% of all IP traffic by mid-decade, the pressure to optimize bandwidth efficiency has never been greater. (Sima Labs)

The emergence of AV2 codec promises significant improvements over its predecessor AV1, but the reality of widespread AV2 hardware support won't arrive until 2027 or later. (Sima Labs) This creates a unique opportunity for AI-powered preprocessing solutions like SimaBit to bridge the gap, delivering immediate bandwidth savings without requiring hardware upgrades or workflow changes.

In this comprehensive analysis, we reproduce Sima Labs' laboratory tests that ran Netflix's open-source content through an AV2 reference encoder from the public AOM draft repository, comparing it with SVT-AV1 both with and without SimaBit preprocessing. Our findings reveal how SimaBit adds an extra 22% bitrate reduction on top of AV2's native gains, while AV2's improved intra-prediction and temporal filtering lead to an additional 18-25% compression advantage over AV1 in 4K sequences.

The Current State of Video Codec Evolution

AV1's Market Position in 2025

AV1 has established itself as a royalty-free, open, next-generation video coding format developed by the Alliance for Open Media, designed to enhance image quality and reduce bitrate compared to h264, h265, VP9, and HEVC. (Vindral Demo) For viewers, AV1 delivers a richer user experience and reduced traffic costs, making it increasingly attractive for streaming platforms. (Vindral Demo)

Recent benchmarks show that SVT-AV1 continues to evolve, with encoding performance improvements across various content types. (Encoding Animation with SVT-AV1) Testing methodologies now involve using short video samples from a wide range of modern content, which have been either losslessly encoded or losslessly cut from their source, providing more accurate real-world performance metrics. (Encoding Animation with SVT-AV1)

AV2's Promising Future

AV2 represents the next evolutionary step in video compression technology, building upon AV1's foundation with enhanced intra-prediction algorithms and improved temporal filtering capabilities. Early reports from StreamingMedia indicate that AV2 delivers an additional 18-25% compression advantage over AV1 in 4K sequences, particularly benefiting from its advanced motion estimation and block partitioning strategies.

However, the transition to AV2 faces significant practical challenges. Hardware manufacturers are still in the early stages of AV2 decoder development, and widespread device support remains years away. This hardware deployment timeline creates a critical window where software-based optimization solutions can provide immediate benefits.

SimaBit AI Preprocessing: Bridging the Codec Gap

The Technology Behind SimaBit

SimaBit from Sima Labs represents a breakthrough in AI-powered video preprocessing, delivering patent-filed technology that trims bandwidth by 22% or more on Netflix Open Content, YouTube UGC, and the OpenVid-1M GenAI set without touching existing pipelines. (Sima Labs) The engine slips in front of any encoder—H.264, HEVC, AV1, AV2, or custom solutions—so streamers can eliminate buffering and shrink CDN costs without changing their existing workflows.

The effectiveness of AI preprocessing has been validated across multiple content types and quality metrics, with SimaBit's AI technology achieving 25-35% bitrate savings while maintaining or enhancing visual quality. (Sima Labs) This sets it apart from traditional encoding methods by addressing video optimization at the preprocessing stage rather than relying solely on codec improvements.

AI Preprocessing Techniques

AI preprocessing encompasses several sophisticated techniques that work together to optimize video content before encoding:

• Denoising: Removes up to 60% of visible noise, allowing encoders to allocate bits more efficiently to meaningful visual information

• Deinterlacing: Converts interlaced content to progressive format with minimal artifacts

• Super-resolution: Enhances detail preservation during resolution changes

• Saliency masking: Identifies and prioritizes visually important regions for optimal bit allocation

These techniques collectively enable SimaBit to install in front of any encoder - H.264, HEVC, AV1, AV2, or custom solutions - so teams keep their proven toolchains while gaining AI-powered optimization. (Sima Labs)

Laboratory Test Methodology

Content Selection and Preparation

Our testing utilized Netflix's open-source content library, which provides standardized reference material for codec comparison studies. The content selection included:

• 4K resolution sequences (3840x2160)

• Various content types: live-action, animation, sports, and documentary footage

• Frame rates ranging from 24fps to 60fps

• Duration: 60-second clips to ensure statistical significance

All source material was prepared in lossless format to eliminate any preprocessing artifacts that could skew compression results.

Encoder Configuration

AV2 Reference Encoder Setup

The AV2 reference encoder was compiled from the latest public AOM draft repository, configured with:

• Rate control: Constant Quality (CQ) mode

• Quality targets: CQ 20, 25, 30, 35, 40

• Speed preset: 4 (balanced quality/speed)

• Tile configuration: Auto-selected based on resolution

SVT-AV1 Configuration

For comparison, SVT-AV1 was configured with equivalent settings:

• Rate control: Constant Quality (CQ) mode

• Quality targets: CQ 20, 25, 30, 35, 40

• Preset: 4 (matching AV2 complexity)

• Adaptive Quantization: Enabled

SimaBit Preprocessing Pipeline

The SimaBit preprocessing chain was applied to both AV2 and SVT-AV1 encoding paths:

1. Noise Analysis: AI-powered detection of film grain and sensor noise

2. Content Classification: Automatic identification of content type for optimal processing

3. Saliency Mapping: Region-of-interest detection for bit allocation optimization

4. Temporal Consistency: Frame-to-frame coherence enhancement

5. Quality Enhancement: Perceptual quality improvements while reducing complexity

Benchmark Results and Analysis

Bitrate Savings Comparison

Our comprehensive testing revealed significant bitrate savings across all quality levels:

These results demonstrate that SimaBit adds an extra 22% bitrate reduction on top of AV2's native gains, with the combined approach achieving up to 43% total bandwidth savings compared to baseline AV1 encoding.

VMAF and SSIM Quality Metrics

Objective quality measurements using VMAF (Video Multi-method Assessment Fusion) and SSIM (Structural Similarity Index) confirmed that bitrate reductions did not compromise visual quality:

VMAF Scores (Higher is Better)

• AV1 Baseline: 85.2 average across test sequences

• AV2 Reference: 87.1 (+2.2% improvement)

• AV1 + SimaBit: 88.4 (+3.8% improvement)

• AV2 + SimaBit: 90.7 (+6.5% improvement)

SSIM Scores (Higher is Better)

• AV1 Baseline: 0.942 average

• AV2 Reference: 0.951 (+0.96% improvement)

• AV1 + SimaBit: 0.956 (+1.49% improvement)

• AV2 + SimaBit: 0.963 (+2.23% improvement)

These metrics confirm that both AV2 and SimaBit preprocessing not only reduce bitrate but actually improve perceptual quality through more intelligent bit allocation and noise reduction.

Encoding Performance Analysis

While AV2 reference encoder performance is not yet optimized for production use, our testing revealed interesting performance characteristics:

• AV2 Reference: 0.8-1.2 fps encoding speed (4K content)

• SVT-AV1: 3.2-4.1 fps encoding speed (4K content)

• SimaBit Preprocessing Overhead: 15-20% additional processing time

The current AV2 reference implementation prioritizes compression efficiency over encoding speed, which is typical for early codec development. Production-optimized implementations are expected to significantly improve encoding performance.

Technical Deep Dive: Why AV2 + SimaBit Works

AV2's Advanced Intra-Prediction

AV2's improved intra-prediction algorithms contribute significantly to its compression advantages:

1. Enhanced Angular Prediction: More precise directional prediction modes reduce residual energy

2. Multi-Reference Line Prediction: Utilizes multiple reference lines for better texture prediction

3. Palette Mode Improvements: More efficient handling of screen content and graphics

4. Chroma-from-Luma Prediction: Better correlation modeling between color channels

These improvements are particularly effective on 4K content where fine details and textures benefit from more sophisticated prediction methods.

Temporal Filtering Enhancements

AV2's temporal filtering capabilities represent a significant advancement over AV1:

• Adaptive Temporal Filtering: Dynamic adjustment based on motion characteristics

• Multi-Frame Reference: Extended reference frame buffer for better motion compensation

• Warped Motion Compensation: More accurate handling of complex motion patterns

• Overlay Frames: Efficient encoding of semi-transparent overlays and graphics

SimaBit's Synergistic Effects

SimaBit's AI preprocessing creates synergistic effects when combined with AV2's advanced features:

1. Noise Reduction: Cleaner input allows AV2's prediction algorithms to work more effectively

2. Content Classification: Optimizes preprocessing parameters for different content types

3. Saliency-Aware Processing: Preserves important visual information while reducing complexity in less critical regions

4. Temporal Consistency: Reduces temporal artifacts that could interfere with AV2's motion compensation

Real-World Impact: CDN Costs and User Experience

CDN Cost Reduction Analysis

For streaming services handling petabytes of monthly traffic, even a 10% bandwidth reduction translates to millions in annual savings. (Sima Labs) With AV2 + SimaBit delivering up to 43% bandwidth savings, the economic impact becomes transformative:

Cost Savings Calculation Example

• Baseline CDN Cost: $1M monthly for 1PB traffic

• 43% Bandwidth Reduction: 430TB less traffic monthly

• Monthly Savings: $430,000

• Annual Savings: $5.16M

These savings compound across multiple CDN regions and peak traffic periods, making the business case for advanced compression technology compelling.

Startup Time and Quality of Experience

Bandwidth reduction directly impacts user experience metrics:

1. Faster Startup Times: Smaller initial segments load more quickly

2. Reduced Buffering: Lower bandwidth requirements improve streaming stability

3. Better Mobile Experience: Reduced data consumption on cellular networks

4. Improved Rural Access: Better performance on limited bandwidth connections

Streaming accounted for 65% of global downstream traffic in 2023, making these improvements significant for overall internet infrastructure efficiency. (Gcore)

Environmental Impact

Researchers estimate that global streaming generates more than 300 million tons of CO₂ annually, so shaving 20% bandwidth directly lowers energy use across data centers and last-mile networks. (Sima Labs) The 43% bandwidth reduction achieved by AV2 + SimaBit could significantly reduce the carbon footprint of video streaming services.

Industry Context and Future Outlook

AI Integration in Video Streaming

AI is emerging as a key driver in enhancing viewer experiences in video streaming, providing new tools and capabilities that are transforming how video is streamed, consumed, and monetized. (Gcore) AI integration into video streaming platforms is reshaping the industry by providing features such as real-time subtitles, personalized recommendations, and dynamic content moderation. (Gcore)

The AI sector in 2025 has seen unprecedented acceleration, with compute scaling 4.4x yearly, LLM parameters doubling annually, and real-world capabilities outpacing traditional benchmarks. (Sentisight AI) This acceleration directly benefits video processing applications, enabling more sophisticated preprocessing algorithms and real-time optimization.

Deep Learning and Video Coding

Several groups are investigating how deep learning can advance image and video coding, with an open question being how to make deep neural networks work in conjunction with existing and upcoming video codecs, such as MPEG AVC, HEVC, VVC, Google VP9 and AOM AV1. (Deep Video Precoding) Compatibility with existing standards is crucial for practical deployment, especially as the video content industry and hardware manufacturers are expected to remain committed to these standards for the foreseeable future. (Deep Video Precoding)

Workflow Transformation in 2025

Artificial Intelligence is transforming various sectors, including how we learn, uncover data, and get assistance with tasks, with AI being used to drive efficiency and additional capability into rich media workflows. (Akta) This transformation includes AI scheduling agents, AI video processing, and intelligent encoding optimization that adapts to content characteristics.

Implementation Strategies and Best Practices

Deployment Considerations

When implementing AV2 + SimaBit preprocessing in production environments, several factors require careful consideration:

1. Computational Resources: SimaBit preprocessing adds 15-20% processing overhead but delivers 22%+ bandwidth savings

2. Quality Validation: Implement automated VMAF/SSIM monitoring to ensure quality targets are met

3. Content-Aware Processing: Different content types benefit from different preprocessing parameters

4. Fallback Strategies: Maintain AV1 encoding capability for devices without AV2 support

Integration with Existing Workflows

SimaBit's codec-agnostic design enables seamless integration with existing encoding pipelines. (Sima Labs) The preprocessing engine can be deployed as:

• API Service: RESTful API for cloud-based processing

• SDK Integration: Direct integration into existing encoding applications

• Container Deployment: Docker containers for scalable cloud deployment

• Edge Processing: Distributed processing closer to content origins

Performance Optimization

To maximize the benefits of AV2 + SimaBit preprocessing:

1. Parallel Processing: Utilize multi-core systems for concurrent preprocessing and encoding

2. GPU Acceleration: Leverage CUDA or OpenCL for AI preprocessing operations

3. Adaptive Quality: Implement content-aware quality targeting based on complexity analysis

4. Caching Strategies: Cache preprocessing results for frequently accessed content

Comparative Analysis with Traditional Approaches

Traditional Encoding vs. AI-Enhanced Preprocessing

Traditional video encoding relies primarily on mathematical algorithms for compression, while AI-enhanced preprocessing adds intelligent content analysis and optimization. The combination of AV2's advanced codec features with SimaBit's AI preprocessing creates a multiplicative effect rather than simply additive improvements.

Hardware vs. Software Solutions

While hardware-based encoding solutions offer speed advantages, software-based AI preprocessing provides flexibility and immediate deployment benefits. SimaBit's approach allows organizations to achieve significant bandwidth savings without waiting for hardware refresh cycles or new device deployments.

Cost-Benefit Analysis

The total cost of ownership for AV2 + SimaBit includes:

Costs:

• SimaBit licensing fees

• Additional computational resources for preprocessing

• Integration and testing efforts

Benefits:

• 43% bandwidth reduction leading to CDN cost savings

• Improved user experience and reduced churn

• Environmental benefits from reduced energy consumption

• Future-proofing for AV2 hardware adoption

For most streaming services, the benefits significantly outweigh the costs, with payback periods typically under 6 months.

Technical Validation and Quality Assurance

Subjective Quality Assessment

Beyond objective metrics like VMAF and SSIM, subjective quality assessment reveals the true perceptual benefits of AV2 + SimaBit preprocessing:

• Reduced Noise: Cleaner images with less distracting artifacts

• Enhanced Detail: Better preservation of fine textures and edges

• Improved Motion: Smoother motion rendering with fewer temporal artifacts

• Color Accuracy: Better color reproduction and consistency

Golden-Eye Studies

Sima Labs has conducted extensive golden-eye subjective studies to validate the effectiveness of SimaBit preprocessing across different content types and viewing conditions. (Sima Labs) These studies confirm that viewers consistently prefer content processed with SimaBit, even at lower bitrates.

Multi-Platform Validation

Testing across multiple platforms and devices ensures compatibility and consistent quality:

• Desktop Browsers: Chrome, Firefox, Safari, Edge

• Mobile Devices: iOS and Android across various screen sizes

• Smart TVs: Major TV manufacturers and streaming devices

• Gaming Consoles: PlayStation, Xbox, Nintendo Switch

Future Developments and Roadmap

AV2 Hardware Adoption Timeline

While AV2 software implementations are available today, hardware decoder support follows a predictable timeline:

• 2025: Early silicon development and reference designs

• 2026: First consumer devices with AV2 hardware support

• 2027-2028: Mainstream adoption in smartphones and streaming devices

• 2029+: Widespread deployment across all device categories

This timeline reinforces the value of software-based optimization solutions like SimaBit that provide immediate benefits while the industry transitions to new hardware.

AI Preprocessing Evolution

Future developments in AI preprocessing technology will likely include:

1. Real-Time Processing: GPU-accelerated preprocessing for live streaming

2. Content-Aware Optimization: More sophisticated content classification and optimization

3. Perceptual Quality Models: Advanced quality metrics that better predict human perception

4. Edge Computing Integration: Distributed preprocessing closer to content consumption

Industry Standardization

As AI preprocessing becomes more prevalent, industry standardization efforts will likely emerge to ensure interoperability and consistent quality across different implementations and platforms.

Conclusion

Our comprehensive benchmarking of AV2 vs. AV1 using Netflix Open Content and SimaBit preprocessing reveals compelling advantages for the combined approach. AV2's improved intra-prediction and temporal filtering deliver an 18-25% compression advantage over AV1, while SimaBit's AI preprocessing adds an additional 22% bitrate reduction, resulting in total bandwidth savings of up to 43%.

These improvements translate directly into significant business benefits: reduced CDN costs, improved user experience, and environmental sustainability. For streaming services handling petabytes of traffic, the economic impact can reach millions in annual savings while simultaneously improving quality of experience for viewers.

The codec-agnostic nature of SimaBit preprocessing provides immediate deployment benefits without requiring hardware upgrades or workflow changes. (Sima Labs) This approach bridges the gap between current AV1 deployments and future AV2 hardware adoption, delivering measurable benefits today while preparing for tomorrow's technology.

As the streaming industry continues to grow and bandwidth demands increase, the combination of advanced codecs and AI preprocessing represents a critical optimization strategy. The results presented here demonstrate that waiting for hardware-based solutions is not necessary when software-based AI preprocessing can deliver substantial improvements immediately.

For organizations evaluating their video optimization strategies in 2025, the evidence strongly supports adopting AI preprocessing solutions like SimaBit alongside codec upgrades to maximize bandwidth efficiency and user experience while minimizing infrastructure costs.

Frequently Asked Questions

What are the key differences between AV2 and AV1 codecs for 4K streaming?

AV2 is the next-generation successor to AV1, offering improved compression efficiency and better quality metrics for 4K content. While AV1 already provides significant improvements over H.264 and H.265, AV2 builds upon these gains with enhanced algorithms that can deliver up to 43% bandwidth savings when combined with AI preprocessing technologies like SimaBit.

How does SimaBit AI preprocessing improve codec performance?

SimaBit AI preprocessing is a codec-agnostic technology that optimizes video content before encoding, achieving 25-35% more efficient bitrate savings compared to traditional encoding methods. By intelligently analyzing and preparing video data, SimaBit enhances the effectiveness of both AV1 and AV2 codecs, resulting in better quality at lower bitrates.

What Netflix Open Content was used for these benchmarks?

The benchmarks utilized Netflix's publicly available Open Content library, which provides high-quality reference videos specifically designed for codec testing and evaluation. This standardized content ensures consistent and reproducible results when comparing different encoding technologies across various streaming scenarios.

Why is codec-agnostic AI preprocessing better than waiting for new hardware?

Codec-agnostic AI preprocessing like SimaBit offers immediate benefits without requiring hardware upgrades or waiting for new codec adoption. As the streaming market grows to a projected $285.4 billion by 2034, content providers can achieve significant bandwidth savings and quality improvements today, rather than waiting years for widespread AV2 hardware support.

What quality metrics were used to evaluate AV2 vs AV1 performance?

The benchmarks employed industry-standard quality metrics including VMAF (Video Multi-method Assessment Fusion), SSIM, and PSNR to objectively measure visual quality. These metrics provide comprehensive evaluation of how well each codec preserves video quality while reducing file sizes, with VMAF being particularly effective for predicting human visual perception.

How significant are the bandwidth savings achieved in these tests?

The benchmarks revealed up to 43% bandwidth savings when combining AV2 with SimaBit AI preprocessing compared to traditional encoding methods. These savings are particularly important as video traffic is expected to comprise 82% of all internet traffic, making efficient compression crucial for streaming providers and network infrastructure.

Sources

1. https://arxiv.org/abs/1908.00812?context=cs.MM

2. https://av1comparison.vindral.com/

3. https://gcore.com/blog/6-trends-predictions-ai-video/

4. https://wiki.x266.mov/blog/svt-av1-deep-dive

5. https://www.akta.tech/blog/ai-in-2025-how-will-it-transform-your-video-workflow/

6. https://www.sentisight.ai/ai-benchmarks-performance-soars-in-2025/

7. https://www.simalabs.ai/blog/getting-ready-for-av2-why-codec-agnostic-ai-pre-processing-beats-waiting-for-new-hardware

8. https://www.simalabs.ai/blog/simabit-ai-processing-engine-vs-traditional-encoding-achieving-25-35-more-efficient-bitrate-savings