Codex Douyin Talking-Head Edit

Problem

Vertical talking-head footage often has dark faces, overexposed backgrounds, repeated speech, captions covering the face, and music overpowering the voice. This skill turns the pre-publish workflow into verifiable steps instead of starting with effects.

How It Works

• Start gate: confirm platform, 9:16, target duration, whether to keep the original voice, and must-keep/must-cut content.
• Source inventory: check resolution, frame rate, audio tracks, face position, safe text areas, lighting, and noise.
• Picture first: correct the image before overlays, using conservative lifts for brightness, contrast, saturation, and local caption backgrounds.
• Semantic cuts: keep the hook, core claim, one or two explanations, and ending; remove repetition, filler, and long pauses; start speed changes at 1.05x-1.15x.
• Animation packaging: use HyperFrames for title, keyword cards, progress lines, and simple metaphor animations; keep titles in upper safe areas and keywords lower, without covering the face.
• Subtitles and music: keep captions short, use a low or medium music bed, and avoid vocals, clipping, or harsh peaks.
• Export self-check: sample beginning, middle, and end frames; verify face brightness, text placement, animation timing, and audio; preserve revisions with versioned filenames.

Boundaries

Best for Douyin-style vertical talking-head shorts. If the platform, duration, or creative direction is unclear, clarify before editing; it is not intended for complex non-speaking narrative edits and should not add unrequested effects or features.