Generate PBR Maps from Seamless PNG Textures with pbr_batch_tool

1) What is this script?

pbr_batch_tool is a Python utility (Python 3 + NumPy, OpenCV) that:

• scans the input directory in/ recursively,
• generates from each seamless PNG texture a full PBR set: Albedo, Height, Normal, Roughness, AO, Metallic, and optional ORM (R=AO, G=Roughness, B=Metallic),
• mirrors the folder structure in out/,
• optionally packs results into one or several ZIP files,
• always uses wrap-padding (BORDER_WRAP) so seamless tiling is preserved.

The script does not “magically guess” a material type – it generates physically plausible derived maps from the input Albedo using classic techniques: height from luminance, normals from gradients, micro-roughness from local contrast, etc.

2) Where to download script?

You can download from Github 

https://github.com/alexmmx/pbr_batch

3) What kind of input image do you need?

You need at least one seamless PNG texture. Ideally, it should be a clean, tileable surface photo or material scan without baked-in lighting.

Requirements for input:

• Seamless: the texture must tile without visible seams at edges.
• Neutral lighting: no strong shadows, highlights, or directional reflections.
• Resolution: 1024–4096 px (1K–4K). Higher resolution = more detail.
• Format: PNG 8-bit per channel (alpha supported). Avoid JPEG for derived maps.
• Color space: sRGB for Albedo; other maps are grayscale or linear RGB.

If your source photo contains highlights or shadows, clean it first to a “flat diffuse” version. Always check seamlessness using Offset in Photoshop/GIMP/Krita.

4) Installation (Linux/Ubuntu)

# Go to the tool folder
cd ~/pbr_batch_tool

# Create and activate a virtual environment
python3 -m venv .venv
source .venv/bin/activate

# Install dependencies (pinned versions)
pip install --upgrade pip
pip install numpy==1.26.4 opencv-python-headless==4.9.0.80

# Verify
python -c "import numpy, cv2; print(numpy.__version__, cv2.__version__)"

5) Folder structure

Put your input PNGs in in/. The script mirrors all subfolders into out/.

in/
 ├─ floor/aged-ceramic-tiles.png
 └─ wall/painted-plaster.png

out/
 ├─ floor/aged-ceramic-tiles_albedo.png
 │                     _normal.png
 │                     _roughness.png
 │                     _ao.png
 │                     _metallic.png
 │                     _ORM.png
 └─ wall/painted-plaster_*.png

6) Running the tool

python3 pbr_batch.py -i ./in -o ./out

python3 pbr_batch.py -i ./in -o ./out --pack-orm --zip-results

python3 pbr_batch.py -i ./in -o ./out --pack-orm --zip-results --zip-per-folder

python3 pbr_batch.py -i ./in -o ./out \
  --normal-strength 4.0 \
  --roughness-contrast 1.3 \
  --ao-strength 1.5 \
  --pack-orm

7) Parameters

• --normal-strength – intensity of normals (default 3.0)
• --roughness-contrast – boost local contrast for roughness (default 1.2)
• --ao-strength – ambient occlusion strength (default 1.0)
• --metallic – constant metallic value (0–1, default 0)
• --pack-orm – export combined ORM map (R=AO, G=Roughness, B=Metallic)
• --skip-height – don’t export a separate height map
• --zip-results – create one big ZIP with results
• --zip-per-folder – also create per-subfolder ZIPs

8) How maps are generated

Height

Computed from perceptual luminance: 0.2126*R + 0.7152*G + 0.0722*B. Normalized to [0..1].

Normal

Gradients via Sobel (with wrap borders). Normal vector scaled by --normal-strength, then mapped to RGB (purple-blue tangent space normal map).

Roughness

Calculated from multi-scale local contrast. Higher contrast → rougher (brighter). Optionally inverted to “glossiness” with --keep-glossy.

AO (Ambient Occlusion)

Approximate cavities by comparing height to blurred neighborhoods (2/4/8 px radii). Lower than surroundings → darker AO.

Metallic

Scalar map: 0 for dielectrics, 1 for metals. Intermediate values possible.

ORM

Combined optimization map: R=AO, G=Roughness, B=Metallic. Useful for Unreal and Unity pipelines.

Seamless

All filters use BORDER_WRAP, so no seams are introduced.

9) From one image to a complete PBR set

1. Prepare a clean seamless PNG texture (no shadows/highlights, 2K–4K recommended).
2. Place it in in/.
3. Run the tool:

   python3 pbr_batch.py -i ./in -o ./out --pack-orm --zip-results

4. Check the output maps: Albedo, Height, Normal, Roughness, AO, Metallic, ORM.
5. Tweak intensities if needed and rerun.
6. Import maps into Blender, Unreal, Unity, etc.

10) Preview in 3D software

Blender

• Albedo → Base Color
• Roughness → Roughness
• Metallic → Metallic
• Normal → Normal Map → Normal
• AO → Multiply with Albedo

Unreal Engine

• Base Color ← Albedo
• Normal ← Normal
• Roughness ← Roughness
• Metallic ← Metallic
• AO ← AO or ORM Red channel

Unity

• Albedo → Base Map
• Normal → Normal Map
• Roughness → Smoothness (invert if needed)
• Metallic → Metallic

11) Quality checks

• Tiling test: Offset the texture by 50% width/height. Seams should be invisible.
• Scaling: Keep surface detail proportional to real-world scale.
• Compression: Don’t use JPEG for Normal/Roughness — use PNG/TIFF to avoid artifacts.

12) Where to get seamless source images

You can use your own tiles or download free seamless PBR textures at AITEXTURED:

• Wood Textures
• Metal Textures
• Rust Textures
• Grass Textures
• Water Textures
• Lava Textures
• 1000+ Basic Materials

Conclusion

pbr_batch_tool transforms a single seamless PNG into a full set of PBR maps: Albedo, Height, Normal, Roughness, AO, Metallic, and ORM. It processes folders recursively, preserves seamless tiling, and exports ready-to-use ZIP archives. Whether you’re working in Blender, Unreal, or Unity, this tool saves hours of manual work and ensures consistent, high-quality materials for your 3D projects.