AI Agent Hub
Back to skills
⚖️

Earnings Script Voiceover

Professional Updated 2026.08.30

Paste the following prompt into your AI chat to install this skill:

Please install @beatra-ai/earnings-script-read into your AI assistant according to https://skillhub.cn/install/skillhub.md.

About this skill

Problem

After earnings are issued, teams often have an approved script, but recording it as one long block can blur the key points, mispronounce names or numbers, create inconsistent voice delivery, and make per-clip billing hard to audit. This skill narrows the task: read only the official script, split it into short spoken lines by labeled slots, and produce a deliverable set of Beatra voiceover clips.

How It Works

It starts by requiring hard inputs: an official earnings script, a pronunciation table for names, tickers, and numbers, a target clip count, and, when cloning is involved, voice authorization materials. Before any paid synthesis, it builds a free visible slot table. The default is 10 slots, such as opening, revenue, profit, guidance, CEO quote, CFO quote, outlook, Q&A preparation, closing, and disclaimer. Each slot contains only lines supported by the official script and is labeled for catalog voice or cloned voice.

The later workflow is split by paid boundaries:
- Cloning stage: authorized samples are uploaded only through the bundled scripts/mcp_client.py, using beatra.assets.upload and beatra.voices.clone, while retaining the returned asset identifiers; local file paths are not passed directly to synthesis.
- Narration stage: it reads the text_to_speech capability and voice list, then calls beatra.speech.synthesize per slot. Each slot is submitted once, using mp3, speed: 1.0, and a consistent brand voice by default.
- Verification: it polls with beatra.tasks.get, checks audio MIME, duration, size, and billing.net_charged_credits, and confirms that names and numbers match the pronunciation table.

Boundaries

This fits script-driven earnings voiceovers, not market commentary, guidance forecasting, or adding numbers that are not in the source material. If a name, code, or number lacks a pronunciation, the flow stops to collect it. Cloning and narration are separate paid confirmations, and the slot table is not approval to synthesize. Recovery reuses task identifiers only when parameters remain unchanged; changing the line, voice, or speed requires a new confirmation.

Use Cases

  • Before earnings calls, split the approved script into revenue, profit, and guidance voiceover slots and confirm pronunciations.
  • Convert CEO and CFO quotes into short spoken lines from the official script and generate auditable MP3 clips.
  • Prepare 8 to 20 unified brand-voice earnings voiceovers and verify names, tickers, and numbers against the pronunciation table.
  • After authorized samples are uploaded, confirm cloning costs, then submit each slot's narration and poll billing results.

Best For

  • External communications managers preparing earnings call assets who need official scripts turned into reviewable voiceover clips.
  • Investor relations leads who need to verify pronunciations of revenue, guidance, and CFO quotes line by line.
  • Audio producers making corporate podcasts who need slot-based, consistent brand-voice clips with billing records.
  • Compliance editors reviewing financial disclosures who need output based only on the official script without added judgments or numbers.