shitsukan-diffusion-kit

KO | EN

gitlite — search

#electron #kubernetes #graphql #ai #generative-ai #anthropic #cpp #hacktoberfest #mcp #kotlin

shitsukan-diffusion-kit

(★ 9)

Free multi-tap optical diffusion emulation for DaVinci Resolve Studio.

File Explorer

ZIP Download

- - HighlightExtract.dctl
  - ProMistVeil.dctl
- - forum-post.md
- - differenz.jpg
  - example-extra-2.jpg
  - filtered.jpg
  - highlight-extract.jpg
  - node-tree.jpg
  - promist-veil.jpg
  - unfiltered.jpg
- - BlackPromist_1.12.8.drx
- Bauanleitung.md
- CHANGELOG.md
- LICENSE
- README.md
- RELEASE_v1.0.2.md
README.md

# Use via CDN

jsDelivr

jsDelivr serves any public GitHub repository as a CDN with zero setup. Pick a version and a file to get a ready-to-paste link and snippet.

Version

File

Link

Example

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

Free multi‑tap optical diffusion emulation for **DaVinci Resolve Studio**.

**Creator:** Stephan Wagner  
**License:** MIT  
**Status:** Stable (tested on macOS Metal, CUDA)

***

## Overview

Shitsukan Diffusion Kit is a compact toolkit that emulates optical diffusion filters (Black Pro‑Mist, Glimmerglass, etc.) inside DaVinci Resolve using a **scene‑linear, multi‑tap bloom** architecture.

Instead of one large blur, the kit builds bloom from **four blur taps at geometrically increasing radii**, each with its own weight. A dedicated veil pass then adds the low‑frequency softness and highlight desaturation associated with classic glass diffusion.

This project is intended for:

- Colorists who want a controllable, node‑based diffusion workflow.  
- Resolve users who prefer to understand and shape the look themselves rather than rely on black‑box plugins.

There are excellent ready‑made commercial tools available. Shitsukan Diffusion Kit deliberately takes the opposite approach: everything is visible, documented, and hackable.

***

## Features

- **Scene‑linear compatible**  
  Designed for DWG/Intermediate, ACES and other scene‑referred workflows. Bloom is added in linear, where additive operations behave more like light.

- **Four‑tap bloom chain**  
  Highlight extraction → 4 blur taps with increasing radii → additive summation. This approximates a heavy‑tailed point‑spread function (tight core, wide wings).

- **Two focused DCTLs**  
  - `HighlightExtract.dctl` — soft‑threshold highlight isolation.  
  - `ProMistVeil.dctl` — per‑pixel veil with black suppression and highlight desaturation.

- **PowerGrade‑friendly**  
  Supplied as a `.drx` so you can drop it into a node tree, tweak it once, then reuse it across shots.

- **Transparent & extensible**  
  All parameters live in the Color page: blur radii, tap weights, thresholds, veil strength. Extend to 5/7 taps, add halation, or retune for different “filter families”.

***

## Node Graph Architecture

Recommended Resolve topology:

```text
01_CST_IN         # into DaVinci Wide Gamut / scene-linear
  -> 02_Base      # base grade
  -> 03_HighlightExtract (DCTL)
      -> Tap1
      -> Tap2
      -> Tap3
      -> Tap4
      -> LayerMixer_Bloom (Add)
  -> LayerMixer_Mix (Add)  # base + bloom
  -> ProMistVeil (DCTL)
  -> CST_OUT      # back to display / output space
```

### Tap weights (default 4‑tap preset)

| Tap | Relative radius | Gain |
| --- | ---------------:| ----:|
| 1   | 1×              | 0.50 |
| 2   | 2×              | 0.35 |
| 3   | 4×              | 0.20 |
| 4   | 8×              | 0.10 |

**Blur radii (HD, by eye):**  
Tap 1 ≈ 0.6–0.8, Tap 2 ≈ 1.3–1.6, Tap 3 ≈ 2.6–3.2, Tap 4 ≈ 5.0–6.5.

Resolve’s blur radius is not pixel‑linear, so treat these as visual ratios, not hard pixel sizes.

***

## Visuals

Place these files in `images/` and the README will render previews:

- `images/node-tree.jpg` – full node graph  
- `images/highlight-extract.jpg` – HighlightExtract settings  
- `images/promist-veil.jpg` – ProMistVeil settings  
- `images/unfiltered.jpg` – base frame  
- `images/filtered.jpg` – frame with Shitsukan Diffusion

```markdown
![Node Tree](https://raw.githubusercontent.com/50asa/shitsukan-diffusion-kit/main/images/node-tree.jpg)

![Highlight Extract](https://raw.githubusercontent.com/50asa/shitsukan-diffusion-kit/main/images/highlight-extract.jpg)

![ProMistVeil](https://raw.githubusercontent.com/50asa/shitsukan-diffusion-kit/main/images/promist-veil.jpg)

| Unfiltered | Filtered |
|-----------|----------|
| ![Unfiltered](https://raw.githubusercontent.com/50asa/shitsukan-diffusion-kit/main/images/unfiltered.jpg) | ![Filtered](https://raw.githubusercontent.com/50asa/shitsukan-diffusion-kit/main/images/filtered.jpg) |
```

***

## Installation

### 1. Install DCTLs

Copy the DCTL files into your Resolve LUT/DCTL folder and restart Resolve (or refresh LUTs).

**macOS**

```text
~/Library/Application Support/DaVinci Resolve/LUT/
```

**Windows**

```text
C:\ProgramData\Blackmagic Design\DaVinci Resolve\Support\LUT\
```

**Linux**

Use your corresponding Resolve LUT directory.

### 2. Import the PowerGrade (optional, recommended)

1. Open the **Color** page.  
2. In the Gallery, right‑click a PowerGrade album → *Import*.  
3. Select `presets/BlackPromist_1.12.8.drx`.

You can now apply the whole node tree to any clip with a double‑click.

### 3. Manual build (if you prefer)

Follow the topology above or the German `Bauanleitung.md` for a step‑by‑step construction of the node graph.

***

## Recommended Settings

### Bloom core & wings

- Tap gains: `0.50 / 0.35 / 0.20 / 0.10`  
- Blur radii (HD): `0.6–0.8 / 1.3–1.6 / 2.6–3.2 / 5.0–6.5`

Start from these, then adjust:

- Lower gains for cleaner, tighter bloom.  
- Higher gains (especially on the wider taps) for more veil and “atmosphere”.

### Black Pro‑Mist‑style character

- **Black Suppress:** `0.80–0.90`  
- **Veil:** `0.03–0.05`  
- **Highlight Desat:** `0.10–0.25`

Keeps shadows comparatively clean while giving strong bloom around highlights.

### Classic / milkier diffusion

- **Black Suppress:** `0.25–0.50`  
- **Veil:** `0.05–0.10`  
- **Highlight Desat:** `0.15–0.35`

Pushes more low‑frequency haze into the image; closer to non‑“Black” diffusion families.

***

## Positioning & Limitations

- **Educational, not calibrated:**  
  The kit is designed as a transparent, experimental implementation. It does **not** attempt per‑filter calibration or per‑stop matching. Treat it as a starting point to build your own “house filter”.

- **Scene‑linear first:**  
  The node tree assumes a scene‑referred working space (DWG/Intermediate, ACEScct, etc.). It will run in log/Rec.709, but response and highlight roll‑off will be less predictable.

- **Radius by eye:**  
  Because Resolve’s blur radius is not a direct pixel value, exact numeric matches across resolutions are not guaranteed. Re‑dial Tap 1 on a reference shot, then scale the other taps relatively.

***

## Repository Layout

```text
shitsukan-diffusion-kit/
├── README.md
├── CHANGELOG.md
├── RELEASE_v1.0.2.md
├── LICENSE
├── Bauanleitung.md
├── dctl/
│   ├── HighlightExtract.dctl
│   └── ProMistVeil.dctl
├── presets/
│   └── BlackPromist_1.12.8.drx
└── images/
    ├── node-tree.jpg
    ├── highlight-extract.jpg
    ├── promist-veil.jpg
    ├── unfiltered.jpg
    └── filtered.jpg
```

***

## Development & Contributions

Ideas for extensions:

- Additional taps (5/7‑tap variants) for even smoother PSF approximation.  
- Separate halation branch (pre‑tone‑map) combined with the existing bloom.  
- Ready‑made presets for different diffusion “families”.

Issues and PRs are welcome. Please include Resolve version, OS, and GPU backend (Metal / CUDA / OpenCL) when reporting bugs.

***

## License

This project is released under the **MIT License**.  
You can use, modify, integrate, and redistribute it in personal or commercial work. Attribution is appreciated but not required.

***

## Credits

**Design & DCTL:** Stephan Wagner  
**Concept:** Multi‑scale bloom as an approximation of heavy‑tailed optical diffusion.  
**Platform:** DaVinci Resolve Studio 19.1+

***

shitsukan-diffusion-kit

File Explorer

# Use via CDN

Was this content helpful?

Recent Feedback

Download README