The Daily Catch Newscast Studio

Use this skill when the user wants source text turned into a short broadcast-style news segment for The Daily Catch, hosted by Claudia Crabtree.

Inputs

• Source text, article text, notes, release notes, or document excerpts.
• Optional target length, tone, audience, or angle.
• Optional authorized voice reference path for audio rendering.

Output

Produce both:

1. A polished newscast script with speaker labels.
2. An audio file when an authorized voice reference and local voice-rendering command are available.

Editorial Style

• Host: Claudia Crabtree.
• Show: The Daily Catch.
• Tone: crisp, warm, informed, and lightly conversational.
• Keep the segment faithful to the source. Do not invent facts, quotes, numbers, affiliations, or claims.
• Convert dense prose into clear broadcast language.
• Prefer a short intro, 2-4 story beats, and a clean sign-off.
• If the source is ambiguous or thin, say so in the script instead of filling gaps.

Workflow

1. Read the source and identify the main news value.

2. Draft a concise segment in this shape:

   Claudia Crabtree: Good evening, and welcome to The Daily Catch.
   Claudia Crabtree: [Lead sentence with the main news.]
   Claudia Crabtree: [Context and why it matters.]
   Claudia Crabtree: [Key details, constraints, or caveats.]
   Claudia Crabtree: [Closing line.]
   

3. Save or pass the final script to scripts/render_newscast.py.

4. If audio is requested, only use a voice reference the user has authorization to use.

Render Script

The bundled renderer can create the transcript, write a spoken-only script, and optionally call the official OmniVoice CLI:

python scripts/render_newscast.py \
  --input source.txt \
  --out-dir ./daily-catch-output

To render audio with OmniVoice, install OmniVoice so omnivoice-infer is on PATH, then pass an authorized voice reference:

python scripts/render_newscast.py \
  --input source.txt \
  --voice /path/to/authorized-voice.mp3 \
  --audio-out ./daily-catch-output/newscast.wav \
  --device mps

If omnivoice-infer is unavailable, still produce the transcript and explain that audio rendering requires installing OmniVoice locally.

Safety Boundary

Do not clone or imitate a real person's voice unless the user states they have permission to use that voice reference. If permission is unclear, produce the script only and ask for an authorized voice reference before rendering audio.