SuperAPI AI Directory

filmcraft-presenter-inset

Design & creation

View repository

Updated 10-23 21:02

About this skill

Narrated video picture-in-picture

ReadConcrete methods and parameters, follow thestep recipesto bind this session's assets, actual returned IDs, and time ranges. The recipes are new-project examples and do not overwrite existing timelines.

Prerequisites and execution

Two videos and narration.wav, at least 4 seconds. 125% is an example of a 64×48 source producing an 80×60 window: actual scale = target window width / source width × 100; do not treat 125 as a universal ratio. Position is in sequence pixels; letterboxing is allowed when the main frame has a different aspect ratio; real audio-video sync and important regions must be recomputed per asset.

Probe the inputs first, then query the current commands. Read return values in batches before replacing references; save and reopen in separate CLI batches, or keep the same MCP editor instance within the host. Seconds, frames, and ticks must not be mixed; do not treat the example integers as universal times.

Checks and delivery

The main frame is on V1, the speaker is independently transformed on V2, and the single approved narration audio track is on A1. Check that the speaker window stays in bounds, does not cover key content, the background protected area is unchanged, and the actual export retains audio.

Deliver the actually existing new project and agreed media; after saving and reopening, re-decode and play-check the export. When only synthetic samples are available, state clearly that real-asset semantics/audio quality and host sync have not yet been accepted. All necessary references and parameters are within this skill; there are no sibling-skill dependencies.