# Chapter Files
DJ mode uses FFmpeg metadata format for chapters, loaded via mpv's `--chapters-file` option.
## Why FFmpeg Format?
CUE files cannot reference HTTP URLs - mpv treats the `FILE` directive as a local path. The solution is:
- Stream audio from HTTP URL as the main source
- Load chapter metadata from a separate FFmpeg-format file
## FFmpeg Chapters Format
```
;FFMETADATA1
[CHAPTER]
TIMEBASE=1/1000
START=1744000
END=3311000
title=Track Name - Artist
[CHAPTER]
TIMEBASE=1/1000
START=3311000
END=4036000
title=Another Track - Another Artist
```
### Fields
| Field | Description |
|-------|-------------|
| `TIMEBASE=1/1000` | Timestamps are in milliseconds |
| `START` | Chapter start time in ms |
| `END` | Chapter end time in ms |
| `title` | Display name (title + artist) |
## Converting from CUE
Use the `cue-to-chapters` script:
```bash
# Basic conversion
cue-to-chapters input.cue > chapters.txt
# With explicit duration for last chapter (in milliseconds)
cue-to-chapters input.cue 8042343 > chapters.txt
```
### What Gets Converted
| CUE Field | FFmpeg Field |
|-----------|--------------|
| `INDEX 01 MM:SS:FF` | `START` (converted to ms) |
| `TITLE` | Combined into `title` |
| `PERFORMER` | Combined into `title` |
### CUE Time Format
CUE uses `MM:SS:FF` where:
- `MM` = minutes
- `SS` = seconds
- `FF` = frames (75 frames per second)
The converter handles this automatically.
## Chapter Storage
Music For Programming chapters are stored in `~/.voicemode/music-for-programming/`:
```
~/.voicemode/music-for-programming/
├── music_for_programming_49-julien_mier.ffmeta # Episode 49
├── music_for_programming_51-mücha.ffmeta # Episode 51
└── ...
```
**Naming Convention:** Chapter files are named to match the RSS MP3 filename exactly (with `.ffmeta` extension instead of `.mp3`). This ensures automatic pairing with audio files.
Use `mfp-rss-helper filename <episode>` to get the correct filename base for any episode.
## Creating Chapters from Scratch
For audio without existing CUE files:
1. **Get tracklist** - Find track listing from source
2. **Download reference tracks** - Get audio for each track
3. **Match timestamps** - Use audio fingerprinting (MFCC correlation)
4. **Generate FFmpeg format** - Create chapter file
Tools for step 3: librosa (Python), chromaprint/fpcalc
## IPC Chapter Commands
Query chapters via raw IPC:
```bash
# List all chapters
echo '{"command": ["get_property", "chapter-list"]}' | socat - /tmp/voicemode-mpv.sock
# Get current chapter metadata
echo '{"command": ["get_property", "chapter-metadata"]}' | socat - /tmp/voicemode-mpv.sock
# Get current chapter index
echo '{"command": ["get_property", "chapter"]}' | socat - /tmp/voicemode-mpv.sock
# Jump to chapter by index
echo '{"command": ["set_property", "chapter", 3]}' | socat - /tmp/voicemode-mpv.sock
```
