tts-voiceover · Text-to-Speech Voiceover

Turns scripts and copy into AI voiceover — spoken delivery, narration, read-aloud audio — with synchronized sentence-level SRT subtitles.

Closed-source cloud TTS (expressive, near-human) is the default when configured; without keys it falls back to edge-tts (flatter, mechanical). The voice track can be mixed with BGM or dropped into video as narration.

Example invocation: "Voice this script with a female narrator and generate a subtitle file alongside."

Full brief

Positioning

tts-voiceover is a text-to-speech dubbing tool: scripts and copy in, AI voiceover out (spoken delivery, narration, read-aloud). It does human voice — never background music.

Core capabilities

Dual engines: with VOICE_PROVIDER + VOICE_API_KEY configured, closed-source cloud TTS is the default (per-sentence synthesis + assembly + sentence-level SRT, expressive and near-human); without keys it falls back to edge-tts (flat, robotic — fallback only); --engine forces either.
Voice choice: default Xiaoxiao (zh-CN-XiaoxiaoNeural, warm female); Xiaoyi (lively young female), Yunxi (clear male), Yunyang (steady broadcast male), Yunjian (powerful male) among common Chinese voices; Cantonese and Taiwan-Mandarin voices available.
Fine-tuning: speech rate / volume / pitch adjustable; long texts via --file to avoid overlong command lines.
Synced subtitles: --subtitle emits a matching SRT file for caption burn-in.
Output formats: mp3 (default), wav / m4a optional.
Post-processing via shared scripts: voice + BGM mixing, loudness normalization to -14 LUFS, adding the voiceover to video as narration.

Workflow

Optional: pick a voice with voices (common Chinese voices with notes; full list needs internet)
speak to synthesize (optional synced SRT)
Optional post-processing: mix / normalize / add to video

Inputs & outputs

Input	Notes
text / file	Required — text to voice or path to a text file (long text via --file)
voice	Optional — voice id (default Xiaoxiao)
rate / volume / pitch	Optional — speed / volume / pitch tweaks
output	Optional — defaults to `outputs/<topic>/{name}.mp3`

Output: voiceover audio (mp3 / wav / m4a) with optional synced SRT, under outputs/<topic>/.

Boundaries with adjacent skills

ai-music: background music / scores (non-vocal or vocal); tts-voiceover (this): human voiceover / narration / read-aloud. They compose: BGM + narration mixed via video_ops.py bgm.

Fit

Short-video narration and talking-head voiceover
Read-aloud content, explainer and knowledge narration
Voiceover jobs that need a synced subtitle file

Before you start

edge-tts calls Microsoft's online service and needs internet access (set a proxy first on intranet environments; the script forwards it automatically).
Closed-source cloud TTS needs VOICE_PROVIDER + VOICE_API_KEY (metered); the config check goes through model_registry.py configured --group voice — never judged by whether env vars print.
With a Profile, preferred voice and speed are read from account settings; without one, the default voice Xiaoxiao at normal speed applies.

tts-voiceover is part of the Aiglade Skill library. Invoke it from the Aiglade chat box in plain language.

← All 155 skills