SuperAPI AI Directory

filmcraft-jcut-dialogue

Design & creation

View repository

Updated 10-23 21:02

About this skill

Interview audio leading J-cut

Readconcrete methods and parameters, perstep recipebind this session's materials, real returned IDs, and time ranges. The recipe is a new-project example and does not overwrite existing timelines.

Prerequisites and execution

A's confirmed speech ends at 1 second, and the 1 second after A is a retainable reaction shot; B's audio and video must share the same source time base. Do not cut off word endings or cause two utterances to unintentionally overlap for the J-cut. The example uses separate silent video and an approved audio track; for mixed AV material, first split/inspect correspondence and sample rate, and do not blindly copy fixed time points.

Probe the input first, then query the current command. Read returned values in batches before replacing references; with standalone CLI, save and reopen in batches, while in a host keep the same MCP editor instance. Seconds, frames, and ticks must not be mixed, and do not treat the example's integers as universal times.

Inspection and delivery

A video 0–2 seconds, B video 2–5 seconds using B source 1–4 seconds; A audio 0–1 seconds, B audio 1–5 seconds using B source 0–4 seconds, so B's audio enters 1 second early while remaining in sync with picture after B appears.

Deliver the actually existing new project and the agreed media; after saving, reopening, and exporting, decode again and check by playback. If only synthetic samples are available, explicitly state that the real material semantics/audio quality and host synchronization have not yet been accepted. The necessary references and parameters are all within this skill, with no sibling skill dependencies.