ai-music-video

# AI Music Video Generator

Create complete music videos: AI music + AI visuals + ffmpeg assembly.

## Quick Start

```
"90년대 보이밴드 풍 한국어 노래 만들어줘" → music only
"발라드 뮤비 만들어줘" → music + slideshow MV
"EDM 뮤비 풀영상으로" → music + video clips MV
"Suno 뮤비로 만들어줘" → Suno native music video
```

## Workflow

### 1. Plan scenes from lyrics/mood
Before generating, create `prompts.json` — array of scene descriptions derived from the song's lyrics, mood, and narrative. 8-12 scenes for a 3-min song.

```json
[
  {"prompt": "Neon-lit city street at night, rain reflections", "type": "image"},
  {"prompt": "Camera slowly panning across a rooftop at sunset", "type": "video"},
  "A lone figure walking through cherry blossoms"
]
```

### 2. Generate music
```bash
bash scripts/suno_music.sh \
  --prompt "가사 또는 설명" \
  --style "90s boy band pop, korean" \
  --title "너만을 원해" \
  --model V4_5ALL --custom \
  --outdir /tmp/mv_project
```

**Options:**
- `--model V4_5ALL` (default), `V5`, `V4_5PLUS`, `V4_5`, `V4`
- `--instrumental` — no vocals
- `--vocal-gender m|f` — vocal gender hint
- `--negative-tags "Heavy Metal, Drums"` — styles to avoid
- `--music-video` — generate Suno native music video (MP4)
- `--dry-run` — cost check only

**Persona (일관된 스타일 유지):**
- `--persona-id ID` — 기존 페르소나 사용 (같은 보컬/스타일로 여러 곡 생성)
- `--create-persona` — 생성된 곡에서 페르소나 생성 → `persona.json` 저장
- `--persona-name "이름"` / `--persona-desc "설명"` / `--persona-style "스타일"`

**Auto features:**
- 🎤 **Timestamped Lyrics**: Non-instrumental tracks automatically fetch lyrics timestamps and save as `{outdir}/lyrics.srt`
- 🎬 **Suno Native MV**: With `--music-video`, Suno generates a visualized MP4 video directly
- 🎭 **Persona**: With `--create-persona`, extracts voice/style identity for reuse

### 3. Generate visuals (custom MV flow)
```bash
bash scripts/gen_visuals.sh \
  --mode slideshow \
  --prompts-file /tmp/mv_project/prompts....