Render a two-host podcast ad from a brand config and dialogue script. Includes the working planner, full-frame and split-screen assembly, captions from measured character timings, brand end card, approved paid-step adapters, and 65 quality-check falsification cases. Use for short conversational ads with two stable hosts in one room.
Install gooseworks: https://gooseworks.ai/creative-agent/install.md # then, in Claude Code, Cursor, or Codex: /gooseworks use the render-podcast-skit skill
This package contains the working two-host podcast pipeline, examples and quality checks. It plans and renders the edit for free. Voice, image and lip-sync generation use the separately installed provider capabilities.
Read the bundled scripts README first. All relative paths in that guide refer to this installed skill folder. Keep the brand config, dialogue script and outputs in the brand project outside the package. Never edit the examples in place. Install the required sibling capabilities and prepare Python, Pillow, requests, ffmpeg, ffprobe, a caption font and the real brand wordmark or font.
Choose the conversation's tone, room, host dynamic and host appearances with the user. Write fresh dialogue for the brand and respect its claim guardrails. Choose an arc from the bundled menu and actually write that turn structure. Match each confirmed voice to its host before generation; save the voice's name and gender with its ID. The voice picker accepts an exported library so it can propose matches without reading account credentials.
Paid commands only print their plan by default. Paid steps run only on an approved order, with both the confirmation and execution flags. Preserve completed audio and timing responses when resuming. Never re-bill a completed beat to recover a missing local file. Motion inserts are unimplemented and fail explicitly if enabled.
Captions follow the voiceover's own character timings. Brand type comes from real files. Review identity, voice fit, room continuity, lip-sync, caption safety and the brand card on the finished video. A preview or package check alone does not establish final creative quality.
Uses the create-vo-elevenlabs, create-image-gpt-image-fal, create-video-fal and
watch skills.
Research supported advertising promises from a specific buyer's context, current product facts, customer language and observed ads. Turn them into static, video or copy ideas; match video ideas to actual recipes and save a product-scoped angle bank for write-video-ad-script. Product-led discovery can use verified facts without reviews. Reference persistence and reach are observations, not proof of performance.
Analyze ad campaign performance data (Google, Meta, LinkedIn) to identify what's working, what's wasting budget, and specific cut/scale/test recommendations. Runs statistical analysis, funnel diagnostics, and multi-channel budget reallocation with specific dollar-amount shift recommendations and scenario modeling.
For paid lead-gen and participant-recruitment ads, replaces vanity CPA with true CAC per qualified lead by joining ad-platform data with downstream funnel events, surfaces tracking gaps, and classifies every creative into Scale / Keep / Investigate / Cut.