ComfyUI-Optical-Realism

Trying to Kill the "Uncanny Valley" of uniform entropy?

Applies a combination of chromatic aberration, depth aware grain and background desaturation, curves to the black/white floor/ceiling, vignette and screen effects.

This node is an attempt at physics-based post-processing that uses a Depth Map to simulate Atmospheric Perspective, Light Transport, and Physical Lens Imperfections. It trys to turns a "generation" into a "photograph."

The Problem

AI images often suffer from the Frequency Distribution Problem. High-frequency details (noise, texture) are distributed equally across the image. In real photography, physics dictates that:

1. Distant objects have lower contrast and lifted blacks (Atmosphere).
2. Bright background light bleeds over foreground edges (Light Wrap).
3. Lenses are not mathematically perfect (Chromatic Aberration & Vignette).
4. Film grain lives in the emulsion, not floating on top of the image (Depth-Aware Grain).

This node does some shit to address all that. Play with it or don't.

Installation

1. Clone this repo into your custom_nodes folder:

   git clone https://github.com/skatardude10/ComfyUI-Optical-Realism.git

2. Restart ComfyUI.

The Workflow

Crucial: This node requires a Depth Map to work its magic. I highly recommend using Depth Anything V2 (specifically the vit_l model for best edge detection).

Wiring it up:

1. Input Image $\rightarrow$ Remove Alpha (included utility) $\rightarrow$ Optical Realism (Image Input)
2. Input Image $\rightarrow$ Depth Anything V2 $\rightarrow$ Invert Image $\rightarrow$ Optical Realism (Depth Input)
   • Note: My script assumes Black = Near, White = Far. If your depth model outputs the opposite, use an Invert Image node.

The Settings

Here is exactly what the sliders do.

💨 Atmospherics

• Atmosphere Enabled: The master switch. Turns on the depth-based physics.
• Haze Strength: How "thick" is the air? Low values = clear winter day. High values = humid rainforest or foggy street.
• Lift Blacks: The secret sauce. Real shadows in the distance aren't pure black (#000000); they are dark atmospheric grey-blue. This lifts the background shadows to separate the subject from the environment.
• Depth Offset: Pushes the "fog curtain" forward or backward. Negative values push it back (clearer foreground), positive values pull it close (macro feel).

📷 Optical Phenomena

• Light Wrap Strength: Simulates "Bloom" or "Halation." It takes bright background light and bleeds it over the edges of the foreground subject. Kills the "cutout sticker" look.
• Chromatic Aberration: Uses Sub-Pixel Sampling (not just resizing) to create a mathematically smooth, infinite-resolution lens fringing effect. Keep this low (0.001 - 0.003) for realism.
• Vignette Intensity: Darkens the corners to mimic a physical lens barrel. Subtle framing that guides the eye to the center.

🎞️ Film Emulation

• Grain Power: Adds texture. Crucially, this is Depth-Aware. The grain is sharp on the focused subject but gets softer/mushier in the blurred background, just like real film.
• Monochrome Grain:
  • True (Default): Simulates Film Stock. Noise affects luminance only.
  • False: Simulates Digital Sensors. Independent RGB noise (Color noise).
• Highlight Roll-off: AI likes to clip bright lights to pure white instantly. This adds a "shoulder" to the highlights, compressing them softly so they look creamy instead of harsh.

Comparison

I am too lazy to get side by side for all of the subtle things this does. I tried to pick sane defaults that make a solid improvement. Play with the sliders... again, or don't! I tried it with the standard plastic skin AI generated images and I will be using this so I share... it takes depth into account to focus on the subject so it's whatever.

Troubleshooting

"RuntimeError: The size of tensor a (4) must match..."

• Your image has an Alpha channel (RGBA).
• Fix: I included a helper node called Remove Alpha (RGBA to RGB). Put this before the Optical Realism node and the Depth node.

Credits

Obvious