A small toolkit and pipeline for applying film simulation to RAW/DNG images, producing both HDR and SDR outputs and optionally saving them in modern formats (AVIF/HEIC) with HDR metadata.

End-to-end pipeline (RAW → Film → Save)

This repository includes a convenient Film Pipeline node and helper nodes that make it easy to run the full workflow from a single node or chain the steps manually.

Key nodes

DNG Image Reader — Load a DNG/RAW file (supports linear HDR output).
Film Simulation V4.1 (HDR Capable) — Apply film simulation in a linear HDR workflow and generate an HDR image plus an SDR preview.
HDR Merge (Gainmap) — Compute a smooth gainmap from an SDR/HDR pair and merge to produce a final HDR image suitable for AVIF/HEIC export.
Save SDR Preview (PNG/JPEG) — Save the SDR preview (performs linear → sRGB conversion when requested).
Save AVIF/HEIC HDR (Native) — Save PQ-encoded AVIF/HEIC using Rec.2020 primaries and optional HDR metadata.
Film Pipeline (RAW → Film → Save) — One-node orchestration that reads RAW, applies film simulation, optionally merges SDR+HDR, and saves SDR/HDR files.

Typical interactive workflows

Manual (node graph):
- DNG Image Reader → Film Simulation V4.1 (HDR Capable) →
  - connect preview_sdr → Save SDR Preview (PNG/JPEG)
  - connect hdr_image → HDR Merge (Gainmap) (optional) → Save AVIF/HEIC HDR (Native)
One-click (single node):
- Film Pipeline (RAW → Film → Save) — set dng_path or supply an image tensor, configure film and merge parameters, and enable save_sdr / save_hdr as needed.

Programmatic usage example

Below is an example of how to call the FilmPipelineNode programmatically (adjust paths and parameters to your setup):

from comfy_film_simulator.film_pipeline import FilmPipelineNode

pipeline = FilmPipelineNode()

hdr_out, sdr_preview, gainmap_vis, info = pipeline.run_pipeline(
    dng_path="/path/to/input.dng",
    image=None,
    film_preset="Kodak Portra 400",
    wb_temperature_K=5600,
    wb_tint=0.0,
    auto_exposure=True,
    exposure=0.0,
    effect_strength=1.0,
    grain_power=1.0,
    halation_power=1.0,
    local_contrast=0.0,
    is_linear_input=True,
    run_merge=True,
    merge_blur_radius=51,
    merge_max_gain=8.0,
    merge_mix=1.0,
    merge_smoothing_mode="guided",
    save_sdr=True,
    sdr_filename_prefix="preview",
    sdr_format="PNG",
    sdr_quality=95,
    apply_srgb_gamma=True,
    save_hdr=True,
    hdr_filename_prefix="hdr",
    hdr_format="AVIF",
    hdr_quality=90,
    ref_white_nits=203.0
)

print(info)

Notes & recommendations

Guided filter (edge-aware smoothing) is used by default for HDR merging when available (OpenCV contrib's ximgproc.guidedFilter). If that package is not available, the pipeline falls back to Gaussian smoothing.
The pipeline expects linear, scene-referred inputs for HDR processing. The SDR preview and Save SDR node will convert to sRGB for display when requested.
For HDR export, install pillow-heif (to enable HEIF/AVIF saving with Pillow). For guided filtering, install opencv-contrib-python to get the ximgproc module if you want edge-aware merging.

Attribution

The HDR merge and pipeline nodes were conceptually inspired by the projects below. This repository does not copy code verbatim from them; the implementations here are original and borrow only conceptual ideas.

https://github.com/jb-jrdn/Hdr-Gainmap
https://github.com/zidage/AlcedoStudio