minimax-music-playlist

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

MiniMax Music Playlist — Personalized Playlist Generator

MiniMax Music Playlist — 个性化歌单生成器

Scan the user's local music libraries to build a taste profile, then generate a personalized playlist using the MiniMax Music API.

Requires:

minimax-music-gen

skill installed at

~/.claude/skills/minimax-music-gen/

扫描用户的本地音乐库构建品味画像，随后通过MiniMax Music API生成个性化歌单。

依赖：需要在

~/.claude/skills/minimax-music-gen/

路径下安装

minimax-music-gen

skill

Workflow Overview

工作流总览

Scan music sources → Build taste profile → Plan playlist → Generate songs → Play & feedback

扫描音乐来源 → 构建品味画像 → 规划歌单 → 生成歌曲 → 播放与反馈

Language Detection

语言检测

Detect the user's language from their message at the start of the session:

Chinese (中文) → Set
```
LANG=zh
```
— all interactions in Chinese, generate Chinese lyrics
English → Set
```
LANG=en
```
— all interactions in English, generate English lyrics

Pass

--lang $LANG

to generation and playback scripts (not scan scripts). Respond to the user in their detected language. Use the matching template below.

会话启动时从用户消息中检测使用语言：

中文 → 设置
```
LANG=zh
```
— 所有交互使用中文，生成中文歌词
英文 → 设置
```
LANG=en
```
— 所有交互使用英文，生成英文歌词

将

--lang $LANG

参数传递给生成和播放脚本（扫描脚本无需传递）。使用检测到的语言回复用户，匹配使用下方对应模板。

Step 1: Scan Music Sources

步骤1：扫描音乐来源

Scan available local music apps. Run whichever scanners are available — the skill works with any combination of sources.

扫描可用的本地音乐应用，运行所有可用的扫描器 —— 该skill支持任意来源组合。

Apple Music

bash

python3 ~/.claude/skills/minimax-music-playlist/scripts/scan_apple_music.py \
  --output /tmp/apple_music_data.json

The script uses

osascript

to query Music.app. It auto-filters TTS/test files. If Music.app is not running or has no library, the script outputs an empty result.

bash

python3 ~/.claude/skills/minimax-music-playlist/scripts/scan_apple_music.py \
  --output /tmp/apple_music_data.json

该脚本使用

osascript

查询Music.app，会自动过滤TTS/测试文件。如果Music.app未运行或无音乐库，脚本将输出空结果。

QQ Music

bash

python3 ~/.claude/skills/minimax-music-playlist/scripts/scan_qq_music.py \
  --output /tmp/qq_music_data.json

Reads QQ Music's SQLite database and preferences plist. Extracts songs, playlist assignments, and search history. If QQ Music is not installed, outputs empty result.

bash

python3 ~/.claude/skills/minimax-music-playlist/scripts/scan_qq_music.py \
  --output /tmp/qq_music_data.json

读取QQ音乐的SQLite数据库和偏好设置plist文件，提取歌曲、歌单归属以及搜索历史。如果未安装QQ音乐，输出空结果。

Spotify

bash

python3 ~/.claude/skills/minimax-music-playlist/scripts/scan_spotify.py \
  --output /tmp/spotify_data.json

Reads Spotify's local LevelDB cache, queries the running app via osascript, and scans the URL cache for API responses. If Spotify is newly installed, data may be sparse — the script warns about this. Data accumulates as the user listens more.

bash

python3 ~/.claude/skills/minimax-music-playlist/scripts/scan_spotify.py \
  --output /tmp/spotify_data.json

读取Spotify的本地LevelDB缓存，通过osascript查询运行中的应用，扫描URL缓存中的API响应。如果是新安装的Spotify，数据可能较为稀疏 —— 脚本会对此发出警告，数据会随着用户使用逐步积累。

NetEase Cloud Music (网易云音乐)

bash

python3 ~/.claude/skills/minimax-music-playlist/scripts/scan_netease.py \
  --output /tmp/netease_data.json

Reads NetEase's webdata JSON files (tracks, FM queue, recommendations), Cache.db, and infers genre preferences from recommended playlist names and scene tags. The app uses a web-based architecture, so data is in

~/Library/Containers/com.netease.163music/Data/Documents/storage/

bash

python3 ~/.claude/skills/minimax-music-playlist/scripts/scan_netease.py \
  --output /tmp/netease_data.json

读取网易云的webdata JSON文件（曲目、FM队列、推荐内容）、Cache.db，并根据推荐歌单名称和场景标签推断风格偏好。该应用使用web架构，数据存储在

~/Library/Containers/com.netease.163music/Data/Documents/storage/

路径下。

Other Music Apps

其他音乐应用

If the user has other apps (e.g., 酷狗, 酷我) not yet supported, tell them:

If LANG=zh:

目前支持扫描 Apple Music、QQ 音乐、Spotify 和网易云音乐。
如果你用其他平台，可以告诉我你喜欢的风格/艺术家，我手动加入画像。

If LANG=en:

Currently supported: Apple Music, QQ Music, Spotify, and NetEase Cloud Music.
If you use other platforms, tell me your preferred styles/artists and I'll add them to your profile manually.

如果用户使用暂不支持的其他应用（如酷狗、酷我），告知用户：

当LANG=zh时：

目前支持扫描 Apple Music、QQ 音乐、Spotify 和网易云音乐。
如果你用其他平台，可以告诉我你喜欢的风格/艺术家，我手动加入画像。

当LANG=en时：

Currently supported: Apple Music, QQ Music, Spotify, and NetEase Cloud Music.
If you use other platforms, tell me your preferred styles/artists and I'll add them to your profile manually.

Manual Input

手动输入

If no music apps are available, or the user wants to supplement:

If LANG=zh:

🎵 告诉我你喜欢的音乐，我来帮你建画像：

💡 比如：
  · "我喜欢周杰伦、陶喆那种华语R&B"
  · "最近在听很多K-pop和日系city pop"
  · "偏好忧郁慢歌，不太听电子舞曲"

If LANG=en:

🎵 Tell me about your music taste, and I'll build a profile:

💡 For example:
  · "I like Jay Chou and David Tao style C-pop R&B"
  · "I've been listening to a lot of K-pop and Japanese city pop"
  · "I prefer melancholy ballads, not into EDM"

如果无可用音乐应用，或用户希望补充信息：

当LANG=zh时：

🎵 告诉我你喜欢的音乐，我来帮你建画像：

💡 比如：
  · "我喜欢周杰伦、陶喆那种华语R&B"
  · "最近在听很多K-pop和日系city pop"
  · "偏好忧郁慢歌，不太听电子舞曲"

当LANG=en时：

🎵 Tell me about your music taste, and I'll build a profile:

💡 For example:
  · "I like Jay Chou and David Tao style C-pop R&B"
  · "I've been listening to a lot of K-pop and Japanese city pop"
  · "I prefer melancholy ballads, not into EDM"

Step 2: Build Taste Profile

步骤2：构建品味画像

Merge all data sources into a unified taste profile:

bash

python3 ~/.claude/skills/minimax-music-playlist/scripts/build_taste_profile.py \
  --apple-music /tmp/apple_music_data.json \
  --qq-music /tmp/qq_music_data.json \
  --spotify /tmp/spotify_data.json \
  --netease /tmp/netease_data.json \
  --gen-history ~/Music/minimax-gen/ \
  --artist-map ~/.claude/skills/minimax-music-playlist/data/artist_genre_map.json \
  --lang $LANG \
  --output ~/.claude/skills/minimax-music-playlist/data/taste_profile.json

The script infers genres from artists using the artist-genre mapping table, computes weighted scores, and outputs a taste profile.

Show the user a privacy-safe summary — never expose raw track lists:

If LANG=zh:

🎵 你的音乐画像：
  📊 数据源：QQ音乐 145首 | Apple Music 42首 | Spotify 20首 | 网易云 15首 | 已生成 9首
  🎸 Top 风格：华语流行 35% | R&B 20% | K-pop 10% | 中国风 8%
  🌍 语言：中文 70% | 韩语 12% | 英语 10% | 日语 8%
  💭 情绪：忧郁 30% | 浪漫 25% | 温暖 20%
  🎤 声线：偏好男声 55% | 女声 40%
  🎵 Top 艺术家：周杰伦、陶喆、孙燕姿、王力宏、方大同

这个画像准确吗？(确认后开始生成歌单，或告诉我要调整什么)

If LANG=en:

🎵 Your Music Profile:
  📊 Sources: QQ Music 145 | Apple Music 42 | Spotify 20 | NetEase 15 | Generated 9
  🎸 Top styles: C-pop 35% | R&B 20% | K-pop 10% | Chinese-style 8%
  🌍 Languages: Chinese 70% | Korean 12% | English 10% | Japanese 8%
  💭 Moods: Melancholy 30% | Romantic 25% | Warm 20%
  🎤 Vocals: Male 55% | Female 40%
  🎵 Top artists: Jay Chou, David Tao, Stefanie Sun, Leehom Wang, Khalil Fong

Does this look right? (Confirm to generate playlist, or tell me what to adjust)

将所有来源的数据合并为统一的品味画像：

bash

python3 ~/.claude/skills/minimax-music-playlist/scripts/build_taste_profile.py \
  --apple-music /tmp/apple_music_data.json \
  --qq-music /tmp/qq_music_data.json \
  --spotify /tmp/spotify_data.json \
  --netease /tmp/netease_data.json \
  --gen-history ~/Music/minimax-gen/ \
  --artist-map ~/.claude/skills/minimax-music-playlist/data/artist_genre_map.json \
  --lang $LANG \
  --output ~/.claude/skills/minimax-music-playlist/data/taste_profile.json

该脚本通过艺术家-风格映射表推断艺术家对应的风格，计算加权得分，最终输出品味画像。

向用户展示隐私安全的汇总信息 —— 绝对不要暴露原始曲目列表：

当LANG=zh时：

🎵 你的音乐画像：
  📊 数据源：QQ音乐 145首 | Apple Music 42首 | Spotify 20首 | 网易云 15首 | 已生成 9首
  🎸 Top 风格：华语流行 35% | R&B 20% | K-pop 10% | 中国风 8%
  🌍 语言：中文 70% | 韩语 12% | 英语 10% | 日语 8%
  💭 情绪：忧郁 30% | 浪漫 25% | 温暖 20%
  🎤 声线：偏好男声 55% | 女声 40%
  🎵 Top 艺术家：周杰伦、陶喆、孙燕姿、王力宏、方大同

这个画像准确吗？(确认后开始生成歌单，或告诉我要调整什么)

当LANG=en时：

🎵 Your Music Profile:
  📊 Sources: QQ Music 145 | Apple Music 42 | Spotify 20 | NetEase 15 | Generated 9
  🎸 Top styles: C-pop 35% | R&B 20% | K-pop 10% | Chinese-style 8%
  🌍 Languages: Chinese 70% | Korean 12% | English 10% | Japanese 8%
  💭 Moods: Melancholy 30% | Romantic 25% | Warm 20%
  🎤 Vocals: Male 55% | Female 40%
  🎵 Top artists: Jay Chou, David Tao, Stefanie Sun, Leehom Wang, Khalil Fong

Does this look right? (Confirm to generate playlist, or tell me what to adjust)

Profile caching

画像缓存

The taste profile is saved at:

~/.claude/skills/minimax-music-playlist/data/taste_profile.json

First time: Always scan and build
Subsequent calls: If a profile exists and is less than 7 days old, reuse it. Show the summary and ask if the user wants to rescan.
Force rescan: If the user says "重新扫描" or "rescan", delete and rebuild.

品味画像存储路径：

~/.claude/skills/minimax-music-playlist/data/taste_profile.json

首次使用：始终执行扫描和构建流程
后续调用：如果画像存在且生成时间不足7天，可复用该画像，展示汇总信息并询问用户是否需要重新扫描
强制重新扫描：如果用户发送"重新扫描"或"rescan"，删除旧画像并重新构建

Step 3: Plan Playlist

步骤3：规划歌单

Ask the user for a theme (optional):

If LANG=zh:

🎵 想要什么主题的歌单？

💡 比如：
  · "深夜放松" — 适合睡前的慢歌
  · "通勤路上" — 节奏轻快提神
  · "雨天" — 配合窗外的雨
  · "让我自己选" — 不指定主题，根据画像随机
  · 或者直接描述你想要的氛围

If LANG=en:

🎵 What theme for your playlist?

💡 Examples:
  · "Late night chill" — relaxing slow songs
  · "Commute" — upbeat and energizing
  · "Rainy day" — matching the rain outside
  · "Surprise me" — random based on your profile
  · Or just describe the vibe you want

If the user doesn't specify, use the top mood from their profile as default.

Generate the playlist plan:

bash

python3 ~/.claude/skills/minimax-music-playlist/scripts/generate_playlist.py \
  --profile ~/.claude/skills/minimax-music-playlist/data/taste_profile.json \
  --theme "<user theme>" \
  --count <number> \
  --lang $LANG \
  --output /tmp/playlist_plan.json

Show the full plan to the user for review:

If LANG=zh:

🎵 歌单计划：深夜放松 (5首)

  1. 华语R&B慢歌 — 忧郁, 内省 [中文/男声]
     Prompt: R&B, neo-soul, 忧郁, 内省, 温柔男声, 电钢琴, 贝斯, 慢板, 深夜独处

  2. K-pop 抒情 — 温暖, 浪漫 [韩语/女声]
     Prompt: K-pop, 流行, 温暖, 浪漫, 清澈女声, 合成器, 钢琴, 中板, 星空下

  3. 独立民谣 — 内省, 平静 [中文/男声]
     Prompt: 独立民谣, folk, 内省, 平静, 温柔男声, 原声吉他, 口琴, 慢板, 深夜独处

  4. Lo-fi hip-hop — 平静, 梦幻 [纯音乐]
     Prompt: lo-fi hip-hop, 平静, 梦幻, 采样钢琴, 电子鼓, vinyl crackle, 慢板, 深夜书桌

  5. 爵士 — 温暖, 浪漫 [英语/女声]
     Prompt: smooth jazz, bossa nova, 温暖, 浪漫, 清澈女声, 钢琴, 贝斯, 萨克斯, 中板

确认生成？(直接回车确认，或告诉我要调整哪首)

If LANG=en:

🎵 Playlist Plan: Late Night Chill (5 songs)

  1. C-pop R&B Ballad — melancholy, introspective [Chinese/male vocal]
     Prompt: R&B, neo-soul, melancholy, introspective, gentle male voice, electric piano, bass, slow tempo, late night solitude

  2. K-pop Ballad — warm, romantic [Korean/female vocal]
     Prompt: K-pop, pop, warm, romantic, clear female voice, synth, piano, mid-tempo, under the stars

  3. Indie Folk — introspective, calm [Chinese/male vocal]
     Prompt: indie folk, folk, introspective, calm, gentle male voice, acoustic guitar, harmonica, slow tempo, late night solitude

  4. Lo-fi hip-hop — calm, dreamy [instrumental]
     Prompt: lo-fi hip-hop, calm, dreamy, sampled piano, electronic drums, vinyl crackle, slow tempo, late night desk

  5. Jazz — warm, romantic [English/female vocal]
     Prompt: smooth jazz, bossa nova, warm, romantic, clear female voice, piano, bass, saxophone, mid-tempo

Confirm? (press enter to confirm, or tell me which song to adjust)

The user can:

Confirm all → proceed to generation
Modify individual songs ("第3首换成摇滚")
Change count ("只要3首")
Regenerate plan ("换一批")

询问用户所需的歌单主题（可选）：

当LANG=zh时：

🎵 想要什么主题的歌单？

💡 比如：
  · "深夜放松" — 适合睡前的慢歌
  · "通勤路上" — 节奏轻快提神
  · "雨天" — 配合窗外的雨
  · "让我自己选" — 不指定主题，根据画像随机
  · 或者直接描述你想要的氛围

当LANG=en时：

🎵 What theme for your playlist?

💡 Examples:
  · "Late night chill" — relaxing slow songs
  · "Commute" — upbeat and energizing
  · "Rainy day" — matching the rain outside
  · "Surprise me" — random based on your profile
  · Or just describe the vibe you want

如果用户未指定主题，默认使用画像中占比最高的情绪作为主题。

生成歌单规划：

bash

python3 ~/.claude/skills/minimax-music-playlist/scripts/generate_playlist.py \
  --profile ~/.claude/skills/minimax-music-playlist/data/taste_profile.json \
  --theme "<user theme>" \
  --count <number> \
  --lang $LANG \
  --output /tmp/playlist_plan.json

向用户展示完整规划供确认：

当LANG=zh时：

🎵 歌单计划：深夜放松 (5首)

  1. 华语R&B慢歌 — 忧郁, 内省 [中文/男声]
     Prompt: R&B, neo-soul, 忧郁, 内省, 温柔男声, 电钢琴, 贝斯, 慢板, 深夜独处

  2. K-pop 抒情 — 温暖, 浪漫 [韩语/女声]
     Prompt: K-pop, 流行, 温暖, 浪漫, 清澈女声, 合成器, 钢琴, 中板, 星空下

  3. 独立民谣 — 内省, 平静 [中文/男声]
     Prompt: 独立民谣, folk, 内省, 平静, 温柔男声, 原声吉他, 口琴, 慢板, 深夜独处

  4. Lo-fi hip-hop — 平静, 梦幻 [纯音乐]
     Prompt: lo-fi hip-hop, 平静, 梦幻, 采样钢琴, 电子鼓, vinyl crackle, 慢板, 深夜书桌

  5. 爵士 — 温暖, 浪漫 [英语/女声]
     Prompt: smooth jazz, bossa nova, 温暖, 浪漫, 清澈女声, 钢琴, 贝斯, 萨克斯, 中板

确认生成？(直接回车确认，或告诉我要调整哪首)

当LANG=en时：

🎵 Playlist Plan: Late Night Chill (5 songs)

  1. C-pop R&B Ballad — melancholy, introspective [Chinese/male vocal]
     Prompt: R&B, neo-soul, melancholy, introspective, gentle male voice, electric piano, bass, slow tempo, late night solitude

  2. K-pop Ballad — warm, romantic [Korean/female vocal]
     Prompt: K-pop, pop, warm, romantic, clear female voice, synth, piano, mid-tempo, under the stars

  3. Indie Folk — introspective, calm [Chinese/male vocal]
     Prompt: indie folk, folk, introspective, calm, gentle male voice, acoustic guitar, harmonica, slow tempo, late night solitude

  4. Lo-fi hip-hop — calm, dreamy [instrumental]
     Prompt: lo-fi hip-hop, calm, dreamy, sampled piano, electronic drums, vinyl crackle, slow tempo, late night desk

  5. Jazz — warm, romantic [English/female vocal]
     Prompt: smooth jazz, bossa nova, warm, romantic, clear female voice, piano, bass, saxophone, mid-tempo

Confirm? (press enter to confirm, or tell me which song to adjust)

用户可以执行以下操作：

全部确认 → 进入生成流程
修改单首歌曲（"第3首换成摇滚"）
更改歌曲数量（"只要3首"）
重新生成规划（"换一批"）

Step 4: Generate & Play (Stream Mode)

步骤4：生成与播放（流模式）

Generate songs sequentially. Play each song as soon as it's ready instead of waiting for all songs to finish. This gives the user music to listen to while the rest of the playlist is being generated.

按顺序生成歌曲，每首歌生成完成后立即播放，无需等待所有歌曲生成完毕。这样用户可以在剩余歌单生成的过程中就听到音乐。

Generation + Playback Loop

生成+播放循环

For each song in the plan:

instrumental: false

, generate lyrics first:

bash

python3 ~/.claude/skills/minimax-music-gen/scripts/generate_lyrics.py \
  --prompt "<lyrics_prompt from plan>" \
  --lang $LANG \
  --output /tmp/playlist_lyrics_<NN>.txt

Generate music (always

--no-play

during generation):

bash

python3 ~/.claude/skills/minimax-music-gen/scripts/generate_music.py \
  --prompt "<prompt from plan>" \
  --lyrics "<lyrics or empty>" \
  --output ~/Music/minimax-gen/playlists/<playlist_name>/<filename> \
  --lang $LANG \
  --no-play

Add --instrumental for instrumental tracks. Omit for vocal tracks.

After each song completes, decide whether to play it immediately:

Song 1 (first song): Always play immediately after generation. Use

play_music.py --background

so playback starts without blocking generation of the next song:

bash

python3 ~/.claude/skills/minimax-music-gen/scripts/play_music.py \
  ~/Music/minimax-gen/playlists/<playlist_name>/<filename> \
  --background --lang $LANG

Song 2+ (subsequent songs): Check if Music.app is idle (not playing). Use the

is_idle

check from

play_playlist.py

bash

python3 -c "
import subprocess
r = subprocess.run(['osascript', '-e',
  'tell application \"System Events\" to (name of processes) contains \"Music\"'],
  capture_output=True, text=True, timeout=5)
if r.stdout.strip() != 'true':
    print('idle')
else:
    s = subprocess.run(['osascript', '-e',
      'tell application \"Music\" to player state as string'],
      capture_output=True, text=True, timeout=5)
    state = s.stdout.strip().lower()
    print('idle' if 'playing' not in state else 'playing')
"

If idle → play this song immediately with
```
play_music.py --background
```
If playing → do NOT interrupt; the song is saved to disk and will be available for replay later. Show a note:

If LANG=zh:

💾 [2/5] 已保存，播放器正忙，稍后可重播

If LANG=en:

💾 [2/5] Saved, player busy — available for replay later

Show progress after each song:

If LANG=zh:

✅ [1/5] 生成完毕：01_bossa_nova.mp3
▶️  正在播放...
⏳ [2/5] 正在生成...

If LANG=en:

✅ [1/5] Complete: 01_bossa_nova.mp3
▶️  Now playing...
⏳ [2/5] Generating...

If generation fails for a song, log the error and continue with the next song.

对规划中的每首歌执行以下操作：

如果

instrumental: false

，首先生成歌词：

bash

python3 ~/.claude/skills/minimax-music-gen/scripts/generate_lyrics.py \
  --prompt "<lyrics_prompt from plan>" \
  --lang $LANG \
  --output /tmp/playlist_lyrics_<NN>.txt

生成音乐（生成过程中始终添加

--no-play

参数）：

bash

python3 ~/.claude/skills/minimax-music-gen/scripts/generate_music.py \
  --prompt "<prompt from plan>" \
  --lyrics "<lyrics or empty>" \
  --output ~/Music/minimax-gen/playlists/<playlist_name>/<filename> \
  --lang $LANG \
  --no-play

纯音轨添加--instrumental参数，有人声的音轨省略该参数

每首歌生成完成后，判断是否立即播放：

第1首歌：生成完成后始终立即播放，使用

play_music.py --background

参数，这样播放启动不会阻塞下一首歌曲的生成：

bash

python3 ~/.claude/skills/minimax-music-gen/scripts/play_music.py \
  ~/Music/minimax-gen/playlists/<playlist_name>/<filename> \
  --background --lang $LANG

第2首及后续歌曲：检查Music.app是否处于空闲状态（未播放），使用

play_playlist.py

中的

is_idle

检测逻辑：

bash

python3 -c "
import subprocess
r = subprocess.run(['osascript', '-e',
  'tell application \"System Events\" to (name of processes) contains \"Music\"'],
  capture_output=True, text=True, timeout=5)
if r.stdout.strip() != 'true':
    print('idle')
else:
    s = subprocess.run(['osascript', '-e',
      'tell application \"Music\" to player state as string'],
      capture_output=True, text=True, timeout=5)
    state = s.stdout.strip().lower()
    print('idle' if 'playing' not in state else 'playing')
"

如果空闲 → 使用
```
play_music.py --background
```
立即播放该歌曲
如果正在播放 → 不要打断播放，歌曲已保存到磁盘，后续可重播，展示提示：

当LANG=zh时：

💾 [2/5] 已保存，播放器正忙，稍后可重播

当LANG=en时：

💾 [2/5] Saved, player busy — available for replay later

每首歌处理完成后展示进度：

当LANG=zh时：

✅ [1/5] 生成完毕：01_bossa_nova.mp3
▶️  正在播放...
⏳ [2/5] 正在生成...

当LANG=en时：

✅ [1/5] Complete: 01_bossa_nova.mp3
▶️  Now playing...
⏳ [2/5] Generating...

如果某首歌生成失败，记录错误并继续处理下一首歌。

When all songs complete:

所有歌曲生成完成后：

If LANG=zh:

🎉 歌单「深夜放松」生成完毕！共 5 首

📁 文件：~/Music/minimax-gen/playlists/深夜放松/
  01_rnb_midnight.mp3
  02_kpop_stars.mp3
  ...

💡 重播整个歌单：告诉我「重播」

If LANG=en:

🎉 Playlist "Late Night Chill" complete! 5 songs

📁 Files: ~/Music/minimax-gen/playlists/late_night_chill/
  01_rnb_midnight.mp3
  02_kpop_stars.mp3
  ...

💡 Replay the full playlist: just say "replay"

当LANG=zh时：

🎉 歌单「深夜放松」生成完毕！共 5 首

📁 文件：~/Music/minimax-gen/playlists/深夜放松/
  01_rnb_midnight.mp3
  02_kpop_stars.mp3
  ...

💡 重播整个歌单：告诉我「重播」

当LANG=en时：

🎉 Playlist "Late Night Chill" complete! 5 songs

📁 Files: ~/Music/minimax-gen/playlists/late_night_chill/
  01_rnb_midnight.mp3
  02_kpop_stars.mp3
  ...

💡 Replay the full playlist: just say "replay"

Replaying existing playlists

重播已生成的歌单

For replaying a previously generated playlist, use the dedicated script:

bash

python3 ~/.claude/skills/minimax-music-gen/scripts/play_playlist.py \
  ~/Music/minimax-gen/playlists/<playlist_name>/ \
  --lang $LANG \
  --auto

如需重播之前生成的歌单，使用专用脚本：

bash

python3 ~/.claude/skills/minimax-music-gen/scripts/play_playlist.py \
  ~/Music/minimax-gen/playlists/<playlist_name>/ \
  --lang $LANG \
  --auto

Step 5: Save Playlist

步骤5：保存歌单

Create a playlist metadata file:

~/Music/minimax-gen/playlists/<playlist_name>/playlist.json

Content:

json

{
  "name": "深夜放松",
  "theme": "深夜放松",
  "created_at": "ISO timestamp",
  "song_count": 5,
  "songs": [
    {
      "index": 1,
      "filename": "01_rnb_midnight.mp3",
      "prompt": "...",
      "lyrics": "...",
      "rating": null
    }
  ]
}

创建歌单元数据文件：

~/Music/minimax-gen/playlists/<playlist_name>/playlist.json

内容：

json

{
  "name": "深夜放松",
  "theme": "深夜放松",
  "created_at": "ISO timestamp",
  "song_count": 5,
  "songs": [
    {
      "index": 1,
      "filename": "01_rnb_midnight.mp3",
      "prompt": "...",
      "lyrics": "...",
      "rating": null
    }
  ]
}

Step 6: Feedback & Profile Update

步骤6：反馈与画像更新

After listening, ask for feedback:

If LANG=zh:

🎵 这个歌单怎么样？

  对每首歌打分（可选）：
  1. 华语R&B慢歌 — 👍 / 👎 / 跳过
  2. K-pop 抒情 — 👍 / 👎 / 跳过
  ...

  或者整体评价：
  🎉 很满意！以后多来这种
  🔄 还行，某些方面可以调整
  🗑️ 不太喜欢，换个方向

If LANG=en:

🎵 How was this playlist?

  Rate each song (optional):
  1. C-pop R&B Ballad — 👍 / 👎 / skip
  2. K-pop Ballad — 👍 / 👎 / skip
  ...

  Or overall rating:
  🎉 Loved it! More like this
  🔄 Decent, but could adjust some things
  🗑️ Not my vibe, try a different direction

Based on feedback:

Per-song ratings: Update playlist.json with ratings. Extract prompt tags from liked/disliked songs and update taste_profile.json's feedback section.
Overall positive: Record the theme + genre combination as successful.
Needs adjustment: Ask what to change, regenerate specific songs.
Negative: Ask what went wrong, note disliked genres/moods in feedback.

用户收听后询问反馈：

当LANG=zh时：

🎵 这个歌单怎么样？

  对每首歌打分（可选）：
  1. 华语R&B慢歌 — 👍 / 👎 / 跳过
  2. K-pop 抒情 — 👍 / 👎 / 跳过
  ...

  或者整体评价：
  🎉 很满意！以后多来这种
  🔄 还行，某些方面可以调整
  🗑️ 不太喜欢，换个方向

当LANG=en时：

🎵 How was this playlist?

  Rate each song (optional):
  1. C-pop R&B Ballad — 👍 / 👎 / skip
  2. K-pop Ballad — 👍 / 👎 / skip
  ...

  Or overall rating:
  🎉 Loved it! More like this
  🔄 Decent, but could adjust some things
  🗑️ Not my vibe, try a different direction

根据反馈执行对应操作：

单首歌曲评分：将评分更新到playlist.json，从喜欢/不喜欢的歌曲中提取prompt标签，更新taste_profile.json的反馈部分
整体好评：记录该主题+风格组合为成功案例
需要调整：询问修改方向，重新生成指定歌曲
差评：询问问题所在，在反馈中标记用户不喜欢的风格/情绪

Replaying Playlists

重播歌单

If the user asks to play a previously generated playlist:

bash

ls ~/Music/minimax-gen/playlists/

Show available playlists and play the selected one song by song.

如果用户要求播放之前生成的歌单：

bash

ls ~/Music/minimax-gen/playlists/

展示可用歌单，逐首播放用户选择的歌单。

Error Handling

错误处理

Error	Action
No music apps installed	Guide to manual input mode
QQ Music DB locked	Suggest closing QQ Music and retrying
Apple Music not running	Try to launch via osascript, or skip
Spotify newly installed	Warn that data is sparse, suggest using app more
NetEase no track data	Data accumulates with usage, suggest browsing/playing
API timeout during generation	Retry once, then skip that song and continue
Generation fails for a song	Log error, continue with next song
No taste profile exists	Run scan first
Profile older than 7 days	Suggest rescan, but allow using old profile

错误	处理措施
未安装任何音乐应用	引导用户使用手动输入模式
QQ音乐数据库被锁定	建议关闭QQ音乐后重试
Apple Music未运行	尝试通过osascript启动，或跳过该来源
新安装的Spotify	警告数据稀疏，建议多使用应用积累数据
网易云无曲目数据	数据会随使用积累，建议浏览/播放音乐
生成过程中API超时	重试一次，如仍失败则跳过该歌曲继续
单首歌曲生成失败	记录错误，继续处理下一首
无可用品味画像	先执行扫描流程
画像生成超过7天	建议重新扫描，但允许使用旧画像

Notes

注意事项

Privacy: Never show raw track lists. Only show aggregated statistics (top genres, language percentages, top artists).
The artist_genre_map.json at
```
data/artist_genre_map.json
```
can be extended manually if the user's library has artists not in the map.
Generated playlists are saved alongside individual songs in
```
~/Music/minimax-gen/playlists/
```
.
The taste profile is persistent across sessions.
All scripts use Python stdlib only — no pip dependencies.

Reuses

generate_lyrics.py

generate_music.py

play_music.py

from

~/.claude/skills/minimax-music-gen/scripts/

隐私：绝对不要展示原始曲目列表，仅展示聚合统计数据（Top风格、语言占比、Top艺术家）
```
data/artist_genre_map.json
```
路径下的艺术家-风格映射表可以手动扩展，如果用户库中的艺术家不在映射表中可自行添加
生成的歌单和单首歌曲一同保存在
```
~/Music/minimax-gen/playlists/
```
路径下
品味画像在不同会话间持久化存储
所有脚本仅使用Python标准库 —— 无pip依赖

复用

~/.claude/skills/minimax-music-gen/scripts/

路径下的

generate_lyrics.py

、

generate_music.py

、

play_music.py

脚本