Netflix has introduced a significant technical enhancement to its streaming pipeline with the rollout of Film Grain Synthesis in AV1. This feature allows the platform to preserve the textured, artistic quality of film grain while delivering substantial improvements in compression efficiency. As more titles begin using this approach, viewers can expect to experience classic films and grain-heavy visuals with greater fidelity and smoother playback, even at lower bitrates.
Why film grain matters in streaming
Film grain is more than a visual remnant of analog filmmaking. It is a deliberate artistic choice that brings depth, texture, and mood to motion pictures. Whether capturing the grit of a noir thriller or the warm tone of a period drama, grain enhances the emotional and cinematic quality of a scene.
However, grain is inherently random and complex, making it difficult for video encoders to compress efficiently. Traditional methods often flatten or discard grain to reduce file size, which can result in a sterile or overly smooth image. With the rise of high-resolution displays and cinematic streaming, this loss of texture becomes more noticeable.
In digital content production, grain is still very much alive. It may originate from camera sensor noise or be added in post-production. The challenge is how to preserve this controlled chaos without bloating the file size or hurting playback performance.
The AV1 approach to film grain
Instead of compressing grain directly, AV1 introduces a smarter alternative through Film Grain Synthesis. The idea is to remove the grain during encoding, model its behavior separately, and then reconstruct it on the viewer’s device during playback.
This involves a multi-step workflow:
- The original video is first denoised to strip away the grain.
- The clean video is compressed using AV1’s efficient codec structure.
- Grain parameters are extracted from the residual difference between the original and denoised versions.
- These parameters are transmitted with the compressed video.
- The player synthesizes grain in real time and overlays it onto the decoded frames.
This method allows high levels of compression without losing the visual authenticity of the original film. It also results in faster streaming, less buffering, and improved playback stability across devices.
Modeling grain pattern and intensity
AV1 uses two separate models to replicate the look and behavior of film grain: pattern and intensity.
Grain pattern is handled using an auto-regressive model. This analyzes how grain evolves across pixels and estimates spatial relationships using coefficients. A noise template is built based on these values, and during playback, small sections are sampled and randomly applied across the image. This ensures that grain appears natural and organic rather than repetitive or artificial.
Grain intensity is controlled by a scaling function. Since grain behaves differently in dark and bright areas, this model adjusts grain strength based on brightness and color. In darker scenes, grain appears more pronounced, while it fades subtly in brighter sections, mimicking real-world film behavior. This scaling function ensures that the restored grain adapts dynamically to the content on screen.
Together, these models allow for a realistic recreation of grain that feels authentic but is computationally lightweight.
Why this matters for streaming quality
In extensive internal tests across hundreds of titles, Netflix observed substantial benefits:
- Bitrates for 1080p and higher resolutions were reduced by an average of 36%.
- For 720p and below, the drop was closer to 10% due to the natural loss of grain during downscaling.
- The visual experience improved, with synthetic grain masking artifacts and preserving texture.
- Frame-by-frame comparisons showed AV1 FGS delivering more coherent visuals than regular AV1, especially in grain-heavy titles.
In one test case, a frame encoded without FGS required over 8000 kbps, while the same frame with FGS used just 2800 kbps, a savings of more than 65%, without noticeable degradation.
The complexity of measuring grain quality
Unlike traditional encoding metrics, AV1 FGS produces outputs that are not pixel-identical to the source. Because the grain is synthesized on the fly, pixel-wise metrics like PSNR and VMAF penalize it, even when the end result looks better.
Human perception plays a key role in how grain is experienced. Synthesized grain can reduce visible compression artifacts by blending them into a more natural texture. This is known as contrast masking and is part of why FGS often looks better even when the numbers disagree.
Netflix’s real-world deployment impact
Netflix rolled out AV1 Film Grain Synthesis at scale, evaluating its effect on performance across real-world playback scenarios. A/B tests showed:
- Lower starting and average bitrates, improving performance on slower connections
- Fewer playback errors and reduced buffering
- Faster time-to-play and more stable resolution playback across sessions
- A slight increase in 4K viewing hours on supported devices, thanks to reduced bitrate demands
Titles like Baahubali 2, The Hot Spot, and Dept. Q now streams with FGS enabled, providing a richer visual experience while using less data.
Advancing the Art of Streaming Film
Film Grain Synthesis in AV1 reflects a larger trend in streaming: smarter compression that respects creative intent. Rather than removing noise as an unwanted element, this approach treats grain as part of the picture’s identity.
As support for AV1 decoding expands across devices, this method opens the door for broader adoption across platforms. It’s not just about saving bandwidth, it’s about preserving the look and feel of cinema in the streaming era.
