At the SIGGRAPH 2025 conference in Vancouver, three engineers from Polyphony Digital delivered a detailed presentation on the visual technology inside Gran Turismo 7, offering a rare look behind-the-curtains to explain why the game’s graphics have such a distinct quality.
Titled “Driving Toward Reality: Physically Based Tone Mapping and Perceptual Fidelity in Gran Turismo 7“, the talk was presented by lead technical artist Kenichiro Yasutomi, lead graphics engineer Kentaro Suzuki, and image processing engineer Hajime Uchimura.
As with some of the other academic presentations delivered by Polyphony Digital engineers at conferences such as CEDEC, it provides a look behind the scenes at how the company’s custom technology actually works.
“Driving Toward Reality” explains how the studio translates real-world lighting and color onto a player’s screen. Although a video of the presentation is not available, the slides can be downloaded directly from Polyphony’s corporate website.
For those seeking a more technical example of the ideas and algorithms presented, the presentation also includes actual sample C++ code from Polyphony Digital published under the MIT License, or keep reading below for a high-level overview and explanation of the talk.
Table of Contents
- Why Gran Turismo Looks Different
- Gran Turismo 7 vs. GT Sport
- Auto-Exposure Based on Human Eyes
- A Unified System
- Future Work (Gran Turismo 8?)
Why Gran Turismo Looks Different
The central challenge in creating realistic graphics is managing the vast difference in brightness, known as dynamic range, between the real world and what a television can display. This process of compressing the range is called tone mapping.
According to the presentation, Polyphony Digital’s goal is to achieve “perceptual realism”, reproducing an image that matches how our eyes actually perceive light rather than just simulating the raw physics.
The engineers highlighted a key distinction between a camera and the human eye: a camera’s exposure is a linear, uniform adjustment to brightness, but human vision uses a complex, non-linear process called adaptation. We perceive relative contrast more than absolute brightness, allowing us to see details in a wide range of conditions.
This principle of modeling the non-linear nature of human vision is a cornerstone of Gran Turismo’s graphics pipeline.
Gran Turismo 7 vs. GT Sport
For Gran Turismo 7, the team developed a new tone mapping system to address issues from its previous technology used in Gran Turismo Sport.
The older system could sometimes cause “hue twisting”, where highly saturated colors would shift their appearance under bright light (for example, red brake lights could appear orange). However, the new “GT7 Tone Mapping” system is a “Color Volume Mapping” operator, meaning it processes brightness and color information together rather than on separate R, G, and B channels.
This method involves a unique blending process:
- An initial image is processed using a per-channel curve, which can introduce a slight, camera-like hue shift.
- A second, color-accurate version is created by converting the image to a Uniform Color Space (UCS). This space separates brightness (luma) from color (chroma), allowing brightness adjustments without distorting the original hue.
- The two versions are then blended together to produce the final image.
This blended approach is designed to maintain color fidelity for important elements like authentic car paints while allowing for a controlled, natural-looking result in highlights.
Auto-Exposure Based on Human Eyes
To handle the game’s dynamic time-of-day and weather, GT7’s auto-exposure system departs from a simple camera model and instead simulates the two primary adaptation mechanisms of the human eye.
- Short-Term Adaptation: This models the quick, limited response of our pupils and neural circuits, which adjust over a few seconds. This handles sudden changes in light, such as entering or exiting a tunnel.
- Long-Term Adaptation: This simulates the much slower (tens of minutes) chemical changes in the eye’s retinal receptors. This wide-ranging adjustment handles gradual transitions, like the sky darkening from day to night.
A significant outcome of this dual system is that the game does not always attempt to create a “perfectly exposed” image. Because the long-term adaptation sets a physiological limit, a bright sunny day is allowed to feel bright, and a dark night scene remains dark, mimicking the natural limits of our own vision.
To ensure stability, this system measures light not just from the player’s view but also from sky dome data and light probes to get a sense of the total environmental lighting.
A Unified System
The presentation concludes that these components are not separate filters but part of a single, integrated pipeline they call “Physically Based Tone Mapping”.
This system, which also includes effects like glare and auto white balance, is built on principles from physics, psychophysics (the study of perception), and psychology to create a consistent and perceptually grounded visual experience across the game’s diverse conditions.
Future Work (Gran Turismo 8?)
Looking ahead, the presentation outlined several key areas for future work across three main categories.
First is modeling, where the team aims to further refine its simulation of human vision by better accounting for the differences between it and camera sensors, especially in low-light conditions where effects like the Purkinje shift occur. They also plan to model the diversity in color perception that exists between individuals.
Second is parameter adjustment, with a goal of creating systems that can automatically adjust for a wide range of display devices and utilize technologies like eye-tracking for real-time tuning in VR environments, suggesting that VR will continue to be a focus for Polyphony Digital moving forward.
Finally, they hope to establish a more quantitative link between objective physical measurements and the subjective visual evaluations of how an image looks and feels to a person.
See more articles on Graphics and Polyphony Digital.
