What problem does it solve? Translating a video's spoken content into another language while preserving each speaker's voice requires coordinating transcription, translation timing, voice cloning, audio mixing, and quality review across many fragile steps. This Skill manages the entire dubbing pipeline as one resumable project, so a failed render or interrupted session can be inspected and resumed instead of restarted. ## Core Features & Use Cases - State-aware project routing: Inspect an existing project directory and automatically resume analysis, translation authoring, rendering, or delivery validation from the correct stage. - Evidence-reconciled transcription: Combine Omni multimodal analysis with mandatory VAD speech detection to author accurate speaker-labeled transcripts with reconciled timestamps. - Speaker-preserving dubbing: Generate reference-guided IndexTTS2 voices per segment, fit speech to time slots with bounded acceleration, mix with separated background audio, and remux while preserving the source video stream. - Use Case: Given a 20-minute Chinese product demo video, produce an English dubbed version where each original speaker's voice identity is preserved, speech fits the original timing, and a QA report flags segments needing manual listening review. ## Quick Start Translate the speech in my video file into English and produce a dubbed version that preserves each speaker's voice.
