Product Launch Talking Video Preview

Problem it solves

When a seller already has approved launch facts and usable still frames, the next bottleneck is often production control: which frame to animate, which line to speak, whether to clone a voice, and how many clips to deliver. This skill narrows that workflow into one supervised pipeline: each authorized still frame becomes one independent talking preview, defaulting to 3 clips and supporting 2–8 clips. It does not concatenate multiple segments into one final video.

How it works

It first validates hard inputs: a host-viewable frame with MIME, width, height, aspect, byte size, and alpha channel; seller-approved facts about the product, launch timing, and visible content; face or voice rights; and delivery count. Then it creates a free slot table listing the frame, spoken line, target duration, and voice strategy. Paid actions are staged:
- beatra.voices.clone: only after the clone card is shown and accepted.
- beatra.speech.synthesize: one narration per slot.
- beatra.videos.animate: one video per confirmed audio segment.
Remote calls go through the bundled scripts/mcp_client.py, with client_request_id and task_id used for reconciliation, polling, and recovery.

Boundaries and cautions

This fits short launch announcements driven by still frames and approved copy. It is not for generic talking avatars, silent product motion, welcome videos, or seeding-style oral content. Files are not consent; do not invent dates, prices, selling points, or faces. If narration exceeds the allowed clip length, split into separate segments instead of concatenating. Final delivery is based on actual bytes, MIME, duration, and billing.net_charged_credits.