Credit Rights Voice

Problem It Solves

A financial service desk may already have a written credit-card benefits table, but still needs to turn it into tagged, auditable voice clips. This skill targets that benefits-table-to-narration workflow, usually delivering 8 to 20 clips and prioritizing card type, annual fee, and lounge coverage. It reads the existing table only and avoids adding approval guarantees, eligibility conclusions, or benefits that are not listed.

How It Works

• Required inputs: the completed benefits table, a pronunciation list for card names or brands, and target count; when no count is specified, plan around 10 clips and confirm voice-clone authorization when cloning is requested.
• Free slot table first: before paid cloning or narration, split the table into short spoken lines and mark whether each slot uses a catalog voice or a cloned voice.
• Clone and narration are separate: clone samples are uploaded only through scripts/mcp_client.py and beatra.assets.upload, keeping the returned artifact ID; narration is generated with beatra.speech.synthesize as mp3, with each input kept around 50,000 characters or fewer and split by sentence.
• Task verification: poll with beatra.tasks.get, then check the actual audio MIME, duration, size, and billing.net_charged_credits instead of treating estimates as final.

Boundaries

This skill is for existing benefits-table narration, not generic voiceover, wealth-product explanations, or digital-human pitching. A local file is not consent; cloning and narration require separate confirmation, and balance issues should be relayed only with the official message and link. All remote Beatra operations go through the bundled MCP client, with no REST/OpenAPI fallback.