Game Script Voice Pack

Problem

Script-kill scripts, indie game dialogue trees, and interactive stories often involve multiple characters, but voice-pack delivery can still break down: character voices are not consistent, line order drifts, a single retake forces whole-pack re-synthesis, and clips lack line-level metadata. @beatra-ai/game-script-voice-pack targets speaker-annotated scripts and provides an auditable dubbing pipeline: each named character keeps a fixed voice, clips are generated in script order, and delivery fields such as character ID, line ID, duration, and artifact remain attached.

How It Works

The core input is a speaker-annotated script body, not just a list of character names; if only character names are provided, it should not invent dialogue. The skill first builds a character table and line ledger, then calls beatra.voices.list to pin each character's voice_id. Before synthesis, it reads live TTS information and confirms the character table, block count, weighted word count, and estimate.

Each voice block is submitted once through the bundled scripts/mcp_client.py with beatra.speech.synthesize, then polled to completion with beatra.tasks.get. Deliverables include sequentially numbered audio clips, real duration, MIME type, URL or artifact, resolved model, and billing.net_charged_credits. If one line changes, only that block is re-run.

Boundaries And Notes

This skill fits multi-character script dubbing, not literary audio drama, single-narrator courses, or one reusable voice-only needs. Voice cloning is only used when the user explicitly authorizes it; the user must confirm ownership or authorization, the sample must pass checks, and funding must be confirmed before beatra.voices.clone. Confirm the package estimate before paid calls; if a response is lost, replay the frozen ID; if balance is insufficient, recharge first and retry.