Converts scripts into expressive multi-segment TTS payloads for CosyVoice and MiniMax models.
npx skills add https://github.com/Jamailar/RedBox --skill tts-director
Please help me install this Agent Skill. Skill: tts-director Source: https://github.com/Jamailar/RedBox/tree/main/desktop/builtin-skills/tts-director Command: npx skills add https://github.com/Jamailar/RedBox --skill tts-director
Flat, monotone text-to-speech output ruins short-video voiceovers, product explainers, and podcast-style narration. This Skill turns a finished script into a single performance-designed TTS request with the right emotion, speed, pitch, pauses, and per-role voices.
Use the tts-director skill to turn this short-video script into an expressive voiceover with emotional segments and natural pauses.
High-intent search queries and answers about installing and using this skill.
Provide the final script and let the skill split it into performance beats, assigning each segment emotion, speed, pitch, and pause markers. It submits one voice.speech request and returns a single merged audio file ready for playback.
CosyVoice uses SSML with rate, pitch, and volume attributes on the speak tag plus a style prompt, while MiniMax uses structured segments with emotion, speed, and pitch fields. The skill detects the model and builds the correct payload automatically.
Yes, the skill identifies the number of roles, selects a distinct voiceId per speaker from available voices, and creates one segment per speaker turn. Boundary pauses between turns are controlled with pauseAfterSeconds values.
SSML applies only to CosyVoice models inside a video-director managed digital-human or VideoRetalk talking-head flow. MiniMax models never use SSML or prompts; they rely on emotion, speed, and pitch fields within ordered segments.
Flat output happens when all segments use identical emotion and speed values with no pauses. The skill's self-check requires each segment to have a justified tone, varied energy across the arc, and intentional punctuation and pause placement.
uzairansaruziExtract transcripts, chapters, speakers, and summaries from YouTube podcasts into markdown notes.
inference-shGenerates talking head videos with AI avatars, lipsync, and voiceover using OmniHuman or PixWorld.
47thtechcornerGenerate multi-scene AI avatar videos with HeyGen v2 API.
video-production-buddyGenerates customizable speech audio from text input via HeyGen's Starfish TTS API.
