Video Narration Draft Replication for Jianying

Problem

The hard part of video replication is rarely just re-cutting footage. It is preserving a template’s editing structure, caption timing, and narration rhythm on supplied source media while avoiding copies of the original narration, music, watermarks, or brand assets. Ad-hoc captioning and TTS often produce broken scene continuity, split dialogue, guessed caption durations, and chopped source lines.

How It Works

The skill runs a staged, auditable pipeline. It starts with a Phase 0 environment preflight for ffmpeg, ffprobe, Python, the ChatCut connector, and the Jianying draft root. It then builds a canonical template map that links template frame intervals to source in/out points, caption start/end frames, audio mode, and BGM behavior. The main sequence is:

• Picture first: rebuild the picture track using the template’s real cut points and verify scene, action, and composition at 0.5s samples and every cut.
• Caption timing lock: after user approval, freeze the template narration-caption start/end intervals; later work may change copy and speech only.
• Copy fitting: rewrite narration inside the locked caption slot, preserving facts, order, and suspense, while keeping TTS speed within 0.9-1.1x.
• Original dialogue: place continuous source passages at 1x and 0dB, without letting narration overlap the dialogue interval.
• Jianying delivery: output separated editable tracks instead of a flattened MP4.

Boundaries

It is best for authorized template replication with supplied source footage and an editable draft requirement. Missing media, unclear caption boundaries, unsupported Jianying builds, or unavailable ChatCut connectors require environment repair or manual review first. It does not export MP4 by default, does not install unauthorized dependencies, and does not assume CapCut and Jianying draft schemas are interchangeable.