# SRT/ — Subtitle files as an independent TC signal

Drop one `.srt` file per episode in this folder. Naming matches the
episode videos:

```
SRT/
├── README.md
├── 01.srt     ← matches Individual Episodes/01.mp4
├── 02.srt     ← matches 02.mp4
└── …
```

## Why we keep SRTs separately

SRTs are produced by subtitle teams independently from the translator's
cue sheet. That makes them a **third, uncorrelated signal** we can use
alongside the cue sheet's TCs and voice-stem onset detection in the
File Prep audit pipeline.

See
[`build-notes/features/file-prep-audit.md`](../../../build-notes/features/file-prep-audit.md)
— specifically **Check 1a — SRT cross-reference** — for how these files
feed into the triple-check.

## Expected format

Standard SubRip:

```
1
00:00:00,980 --> 00:00:01,939
Residência Imperial Garden
Banheiro

2
00:00:03,480 --> 00:00:04,239
Rafael Silva
Presidente do Grupo Silva
```

Timestamps use `HH:MM:SS,mmm` (comma before ms — SubRip convention).
UTF-8 or cp1252 encoding auto-detected.

## Running the comparison

From the repo root:

```bash
python3 file-prep/srt_compare.py
```

Outputs a human-readable summary to stdout + full report to
`file-prep/reports/srt_compare_YYYYMMDD_HHMMSS.{json,md}`.

## Git

`.srt` files are **tracked** — they're small text, and they're a
source-of-truth document we want versioned alongside the project.