Recapo Talking Head Clip Editor

Problem to Solve

Long talking-head videos often contain padding, repetition, and dead air. Manually picking highlights, trimming timecodes, aligning subtitles, and adding BGM is slow. This skill standardizes the workflow for splitting long voiceover videos into multiple themed short clips, reducing issues such as incomplete takes, misaligned subtitles, and messy background voice.

How It Works

It follows Recapo's voiceover-editing workflow rather than simple silence detection, keyword extraction, or equal time slicing. Key steps include:
- Parameter convergence: read references/parameters.md, confirm missing required items in one pass, and do not silently choose subtitles, aspect ratio, BGM, or themes unless defaults are requested.
- Understanding and selection: after preprocessing, merge ASR results with visual content, then use WithoutNarrationAgent to generate a clip plan based on visual context, target length, output count, and user requirements.
- Synchronized trimming and subtitles: trim original video and ASR audio with the same intervals, then re-run ASR on the complete trimmed audio to build subtitle timing.
- Audio and finish synthesis: optionally use MediaKit cloud voice/background separation, then merge voice clips, synthesize audio, sync speed to target duration, and compose aspect ratio, BGM, and loudness.
- Quality checks: validate final clip count, duration, resolution, and subtitle results; FFmpeg/ffprobe are used to verify encoding capabilities.

Boundaries

It fits long talking-head, interview, or explainer videos where you need two or three differentiated short clips by theme. Multiple outputs are not mechanical copies of the same clip and should include theme or differentiation requirements. File, format, parameter, or credential errors are not blindly retried, while dependency or Windows app-control issues should be fixed per guidance. First use surfaces the Recapo web version, but local execution continues if the user explicitly chooses it.