Plate

A VFX plate-preparation workflow built on FFmpeg — inspect footage, set IN/OUT, and export EXR sequences + proxies in one step.

Plate's GUI — load a clip, scrub through frames, set IN/OUT on the timeline, and export.

---

What it does

Plate orchestrates FFprobe, FFmpeg, and (optionally) PySide6 into a single plate-prep pipeline. Open a clip, choose your frame range, and get back:

• A proxy — lightweight H.264 for review
• An EXR sequence — ready for compositing
• A manifest.json — machine-readable metadata

All in one shot-named folder.

---

Quick start

# Prerequisites: FFmpeg on your PATH, Python 3.10+
pip install plateprep

# CLI — single clip
plate footage.mov --in 1204 --out 1389 --start-frame 1001 --output ./shots

# GUI
pip install "plateprep[gui]"
plate-gui

CLI example output

shots/
└── footage/
    ├── proxy.mp4
    ├── exr/
    │   ├── footage.001204.exr
    │   ├── footage.001205.exr
    │   └── ...
    └── manifest.json

---

Features

• Visual frame-range selection — GUI timeline with draggable IN/OUT handles
• EXR export — 32-bit float, configurable compression (none/rle/zip1/zip16)
• Proxy generation — H.264, configurable max width, burn-in (frame number, timecode, source name)
• Color management — LUT (.cube) or OCIO colorspace transform baked into EXRs and proxies
• ComfyUI export — display-referred 16-bit PNG sequence (OCIO display/view baked in, scaled to a max width) ready for ComfyUI
• Shot scaffolding — --shot img01_env lays out ref/ comfy/ paint/ nuke/ renders/ breakdown/ plates/ task folders with versioned artifact names (img01_env_v001, auto-incremented per export); --new-shot scaffolds an empty shot
• Batch processing — process multiple clips from a JSON file
• Export presets — "ACES 2K", "Rec709 HD", "Archival 4K" with user-defined presets
• Nuke script export — auto-generate a .nk script pointing at your EXR sequence
• Shot queue — queue multiple exports from the GUI, run them in the background

The timeline ruler with IN/OUT handles and frame-accurate scrubbing.

---

CLI reference

plate SOURCE --in IN --out OUT [options]
plate --batch BATCH_FILE [options]

Option	Default	Description
--in	required	IN frame (inclusive)
--out	required	OUT frame (inclusive)
--start-frame	0	First frame number in the source
--output	./output	Root output directory
--proxy-width	1920	Max proxy width (never scales up)
--exr-pixfmt	gbrpf32le	EXR pixel format
--exr-codec	zip1	EXR compression (none/rle/zip1/zip16)
--frame-padding	6	Zero-padding for EXR frame numbers
--skip-exr	off	Skip EXR generation
--skip-proxy	off	Skip proxy generation
--nuke-script	off	Generate a Nuke script
--preset	—	Export preset name
--burn-in	—	Overlays on proxy: frame_number, timecode, source_name
--lut	—	Path to .cube LUT
--ocio-config	—	OCIO config path
--ocio-src	—	Source colorspace
--ocio-dst	—	Destination colorspace
--shot	—	Shot name (e.g. img01_env); switches to the shot layout with versioned names
--shot-version	auto	Force a version number (requires --shot)
--new-shot	—	Scaffold an empty shot folder structure and exit
--comfy	off	Also export display-referred 16-bit PNGs to comfy/
--comfy-width	1024	Max width for ComfyUI PNGs (never scales up)
--ocio-display	—	OCIO display for the ComfyUI export
--ocio-view	—	OCIO view for the ComfyUI export

--comfy bakes --ocio-display/--ocio-view (with --ocio-config + --ocio-src) into the PNGs; without display/view it reuses the main color transform. Given without --ocio-dst, the OCIO options apply to the ComfyUI PNGs only and the EXRs/proxy are left untouched:

plate clip.mov --in 1001 --out 1100 --comfy \
  --ocio-config config.ocio --ocio-src "ACEScct" \
  --ocio-display "sRGB - Display" --ocio-view "ACES 1.0 - SDR Video"

Shot mode

# Scaffold an empty shot
plate --new-shot img01_env --output ./shots

# Export into it — picks the next free version automatically
plate clip.mov --in 1001 --out 1100 --shot img01_env --output ./shots --comfy --nuke-script

shots/
└── img01_env/
    ├── ref/  paint/  renders/  breakdown/
    ├── plates/img01_env_v001/img01_env_v001.001001.exr …
    ├── comfy/img01_env_v001/img01_env_v001.001001.png …
    ├── nuke/img01_env_v001.nk
    ├── proxy_v001.mp4
    └── manifest.json

Re-running the same command writes _v002, and --shot-version 1 overwrites v001. The folder set is configurable via [project] folders = [...] in ~/.plate/config.toml. Batch entries accept per-entry "shot" / "shot_version" keys.

Batch mode

[
  {"source": "clip_a.mov", "in": 1001, "out": 1100},
  {"source": "clip_b.mov", "in": 1204, "out": 1389, "proxy_width": 1280}
]

plate --batch jobs.json --output ./shots

---

manifest.json

Every export produces a manifest.json with shot metadata — designed to plug into downstream automation or AI-assisted ingest pipelines.

{
  "shot": "footage",
  "source": "footage.mov",
  "in": 1204,
  "out": 1389,
  "exported_frames": 186,
  "proxy": "shots/footage/proxy.mp4",
  "exr_dir": "shots/footage/exr",
  "fps": 23.976,
  "width": 4096,
  "height": 2160,
  "colorspace": "bt709",
  "has_audio": true
}

---

Using it as a library

from plate.pipeline import PlatePipeline

pipeline = PlatePipeline(
    source="footage.mov",
    in_frame=1204,
    out_frame=1389,
    start_frame=1001,
    output_root="./shots",
)
result = pipeline.run()
print(result.manifest_path)

---

Requirements

• Python 3.10+
• FFmpeg (provides both ffmpeg and ffprobe) on your PATH
• PySide6 — only for the GUI (pip install plateprep[gui])
• OpenColorIO — only for OCIO transforms (pip install plateprep[ocio])

---

License

MIT