Why FilmGraph?

The film industry relies on dozens of incompatible formats — CCSLs in Word documents, Dialogbücher, SRT subtitles, EDLs, XML timelines, proprietary pipeline outputs. Building tools that work across these formats means spending more time parsing than analyzing.

FilmGraph is a single JSON schema that captures shots, scenes, dialogue, cinematography, characters, locations, and narrative events. Zero dependencies beyond Pydantic. One format for your entire pipeline.

{
  "schema_version": "filmgraph/v1",
  "meta": { "title": "Lost Killers", "duration": 5731.2 },
  "scenes": [{
    "id": 1, "title": "Opening", "start_time": 0, "end_time": 142.5,
    "shots": [{
      "id": "sh-001", "order": 1, "start_time": 0, "end_time": 8.3,
      "cinematography": { "shot_size": "WS", "camera_movement": "static" },
      "visual": { "description": "Wide establishing shot of Berlin skyline at dusk" },
      "audio": { "dialogue": [{ "speaker": "narrator", "text": "Berlin, 1998.", "start_time": 2.1, "end_time": 4.5 }] }
    }]
  }]
}

Install

pip install filmgraph

With importers for .docx files (CCSL, Dialogbuch, AD scripts):

pip install filmgraph[docx]

With OpenTimelineIO support (EDL, FCP XML, AAF):

pip install filmgraph[otio]

Quick Start

from pathlib import Path
from filmgraph import FilmGraph, Scene, Shot, ShotSize, CameraMovement

# Load
fg = FilmGraph.from_json(Path("movie.filmgraph.json").read_text())

# Traverse
for scene in fg.scenes:
    print(f"{scene.title} ({scene.shot_count} shots)")
    for shot in scene.shots:
        print(f"  {shot.cinematography.shot_size}  {shot.visual.description}")

# Save
Path("movie.filmgraph.json").write_text(fg.to_json(indent=2))

Schema

FilmGraph
├── meta: FilmMeta            # title, duration, resolution, data sources
├── entities: Entities         # characters, locations (referenced by ID)
├── scenes: [Scene]            # narrative segments
│   ├── shots: [Shot]          # atomic visual units (cut to cut)
│   │   ├── cinematography     # shot_size, camera_movement, camera_angle
│   │   ├── visual             # description, colors, lighting
│   │   ├── audio              # dialogue lines, music cues, SFX
│   │   └── editorial          # transitions, confidence, thumbnail
│   └── dialogue: [DialogueLine]
└── events: [Event]            # plot turning points

Controlled Vocabularies

Every categorical field uses a controlled enum with an OTHER fallback for extensibility.

Importers

Convert industry formats into FilmGraph with a single function call:

from filmgraph.importers.ccsl import ccsl_to_filmgraph
from filmgraph.importers.srt import srt_to_filmgraph

fg = ccsl_to_filmgraph("FTR3_BloodyTennis.docx")
fg = srt_to_filmgraph("movie.srt", title="My Film")

Each importer also works as a CLI:

python -m filmgraph.importers.ccsl input.docx -o output.filmgraph.json
python -m filmgraph.importers.srt movie.srt -o subs.filmgraph.json

Exporters

Go the other direction — render a FilmGraph as the industry format of your choice:

from filmgraph.exporters.srt import filmgraph_to_srt
from filmgraph.exporters.ccsl import filmgraph_to_ccsl

Path("movie.srt").write_text(filmgraph_to_srt(fg))
filmgraph_to_ccsl(fg, "movie_ccsl.docx")

Every exporter also works as a CLI:

python -m filmgraph.exporters.srt movie.filmgraph.json -o movie.srt
python -m filmgraph.exporters.ccsl movie.filmgraph.json -o movie_ccsl.docx
python -m filmgraph.exporters.otio_export movie.filmgraph.json -o edit.edl
python -m filmgraph.exporters.markdown movie.filmgraph.json -o report.md

Build a FilmGraph Programmatically

from filmgraph import *

fg = FilmGraph(
    meta=FilmMeta(title="My Film", duration=5400.0),
    entities=Entities(characters=[
        Character(id="alice", name="Alice"),
        Character(id="bob", name="Bob"),
    ]),
    scenes=[
        Scene(
            id=1,
            title="Opening",
            summary="Alice meets Bob",
            start_time=0,
            end_time=120,
            shots=[
                Shot(
                    id="sh-001",
                    order=1,
                    start_time=0,
                    end_time=8.5,
                    cinematography=Cinematography(
                        shot_size=ShotSize.WS,
                        camera_movement=CameraMovement.STATIC,
                    ),
                    visual=Visual(description="Wide shot of a café"),
                    audio=Audio(dialogue=[
                        DialogueLine(
                            speaker="alice",
                            text="Is this seat taken?",
                            start_time=3.0,
                            end_time=5.2,
                        ),
                    ]),
                ),
            ],
        ),
    ],
)

print(fg.to_json(indent=2))

JSON Schema

Generate the full JSON Schema for validation or code generation:

import json
from filmgraph import FilmGraph

schema = FilmGraph.model_json_schema()
print(json.dumps(schema, indent=2))

Contributing

Contributions welcome. Please open an issue first to discuss what you'd like to change.

License

MIT · Made by Chapter41