📮 商务合作 / Business：wow.vip@foxmail.com 🎬 抖音 / Douyin：v.douyin.com/07EjIDGVErY
把配音「平、假、没情绪层次」拆成可执行的副语言指令：基于副语言学理论，把语速、音高、音量、停顿、声音质感、非言语发声六大维度，翻译成 TTS 与视频模型听得懂的声音表演提示。先用诊断步骤定位声音到底哪里不对，再用「角色声音基线四维」与「身份-声音映射」给每个角色定一条专属声音线。写指令前先过铁律与反模式清单，排掉模型根本执行不了的空话。它是听觉表演层，与视觉表演层（微表情、眼神、肌肉链）叠加，才凑齐视听同步的角色真实感。
- 声音诊断：先判断声音哪里不对，再决定改什么
- 角色声音基线四维 + 身份-声音映射，一人一条声音线
- 六大维度写指令：语速、音高、音量、停顿、质感、非言语发声
- 声音质感编码：气声、颤声、沙哑、鼻音对应何种情绪
- 非言语发声逐条描述：叹气、哼声、吞咽、清嗓、笑声
- 铁律 + 反模式清单：写指令前先排掉模型执行不了的写法
voice-performance-layer—— 见voice-performance-layer/SKILL.md
git clone https://github.com/yzsp-studio/voice-performance-layer.git
cp -r voice-performance-layer/* "$DSH_HOME/skills/"技能是纯文本说明（
SKILL.md+references/），零依赖、零 API key，用你当前 agent 的会话额度。
Turns flat, fake, emotionless AI dubbing into executable direction: grounded in paralanguage theory, it translates six dimensions — tempo, pitch, volume, pause, voice quality and non-verbal vocalizations — into voice-performance prompts that TTS and video models can actually follow. A diagnosis step first locates what is wrong with the voice, then a four-dimension character voice baseline plus an identity-to-voice mapping gives every character one consistent vocal line. Hard rules and an anti-pattern list come before writing, so you stop producing directions the model cannot execute. It is the auditory performance layer; stacked with the visual performance layer (micro-expressions, eye line, muscle chains) it completes lip-and-ear-synced character believability.
- Voice diagnosis: find what is actually wrong before changing anything
- Four-dimension character voice baseline plus identity-to-voice mapping: one vocal line per character
- Six-dimension direction: tempo, pitch, volume, pause, voice quality, non-verbal vocalizations
- Voice-quality coding: breathy, tremulous, raspy and nasal mapped to specific emotions
- Non-verbal vocalizations described one by one: sighs, hums, swallows, throat clears, laughter
- Hard rules and an anti-pattern list that kills directions the model cannot execute
voice-performance-layer—voice-performance-layer/SKILL.md
git clone https://github.com/yzsp-studio/voice-performance-layer.git
cp -r voice-performance-layer/* "$DSH_HOME/skills/"Skills are plain-text instructions (
SKILL.md+references/). Zero dependencies, no API keys — they run on your agent's existing session quota.
| 🎬 抖音 / Douyin | https://v.douyin.com/07EjIDGVErY/ |
| 📮 商务合作 / Business | wow.vip@foxmail.com |
| 🐙 GitHub | https://github.com/yzsp-ai |
| 💖 打赏 / Support | afdian.com/a/yzsp-ai |