dialogue-editing-adr

What it does

Directs an AI agent to treat dialogue as a performance and editorial contract, focusing on preserving intelligibility, intent, continuity, and rights. Applies to spoken audio in generated videos, explainers, localization, podcasts, and related content. The skill analyzes the type of dialogue problem (production-edit, performance-fix, ADR-replace, localization-dub, VO-style-dub, synthetic-rerender, caption/transcript-handoff) and prescribes specific workflows accordingly. It mandates adherence to industry guidance for loudness, captions, and dubbing practices, and it provides detailed steps for creating a dialogue map, preserving original performance, and performing repairs before considering ADR or synthetic regeneration. It also covers ADR cue preparation, directing performance, capture of multiple takes, maintaining voice continuity, and synthetic voice boundaries, plus handoffs for captions, transcripts, and mix. The guidance emphasizes avoiding broad processing, preserving room tone, and ensuring picture-sync considerations, especially where lip-sync and mouth shapes matter. It lists documented facts on loudness standards (ITU-R BS.1770-5, EBU R 128), caption quality (FCC guidance, W3C WCAG 2.2), Netflix dubbing and mixing principles, iZotope tool caveats, and AI-rights considerations from U.S. copyright and industry bodies. It prescribes empirical checks (headphones, various speaker passes, picture-sync, context in full mix) and production heuristics (build a dialogue map, preserve best original performance, continuous air for dialogue, targeted repairs). It also specifies how to prepare ADR cues with scene context, timecode, exact lines, and synchronization anchors, and how to provide a clean, organized handoff with loudness targets, stems, and notes.

How it works

• Classifies each line or section before touching it into categories: production-edit, performance-fix, ADR-replace, localization-dub, VO-style-dub, synthetic-rerender, or caption/transcript-handoff.
• Recommends adhering to documented facts (loudness standards, captioning requirements, lip-sync considerations, and synthetic-voice boundaries).
• Guides empirical checks on the media (clicks, plosives, room tone, intelligibility, picture-sync, full mix context).
• Advises building a dialogue map with fields: timecodes, speaker, screen status, heard/intended text, issue, treatment, pronunciation notes, rights/consent, and handoff notes, with example shapes.
• Encourages preserving the best original performance using a hierarchy of fixes from alternate mic/channel to ADR or synthetic regeneration only if necessary for intelligibility or rights.
• Provides ADR and dubbing directions: cue preparation with scene context, timecode, exact lines, reasons, performance intent, sync anchors, pronunciation guides, mic target, room targets, and takes requested; and instructs directing performance with actionable verbs and context, plus localization guidance.
• Outlines continuity documentation for recurring voices and strict boundaries against cloning without rights; details synthetic-voice constraints and example direction notes.
• Requires captions/transcripts align with shipped audio and provides requirements for capture and formatting in handoffs.
• Specifies mix handoff contents, loudness measurements, and non-destructive workflow preferences.

When to use it

Use when dialogue must carry story, brand, instruction, character, or accessibility. Apply when the goal is to preserve intelligibility, intent, continuity, and rights while coordinating with mixers, captioners, localization teams, or downstream agents. Trigger according to the classification of the dialogue problem (production-edit, performance-fix, ADR-replace, localization-dub, etc.).

What it can touch

• Tools: claude-code, codex, copilot, cursor.
• It prescribes documenting takes, timecodes, and synchronization anchors, and collecting notes for captions, transcripts, and mix handoffs.

Caveats

• Includes references to standards and guidance (ITU-R BS.1770-5, EBU R 128, FCC, W3C WCAG 2.2, Netflix guides, iZotope RX, U.S. Copyright Office, SAG-AFTRA, NAVA) but does not guarantee outcomes.
• Emphasizes empirical listening tests and project-specific rights clearances; not a universal guarantee of quality or legality.
• Warns about potential audio-processing trade-offs (e.g., de-essing, de-plosive strengths) when using tools.