What problem does it solve? Changing the background, lighting, outfit, or mood of an existing video normally requires reshoots or manual VFX work, and generic AI video tools tend to alter the person's face, drop held objects, or lose the original audio. This Skill restyles a video from a free-form prompt while locking the original identity, motion, speech, and soundtrack. ## Core Features & Use Cases - Prompt-driven scene editing: Edits the first usable frame with gpt-image-2, then propagates that look across the clip with Kling reference-video using the original clip as the identity anchor. - Identity and prop preservation: Locked prompt clauses keep the face, pose, and held objects (phones, products, drinks) intact through every re-render. - Long-clip segmentation and audio repair: Splits sources over 15 seconds into 14.8s windows, concatenates segments, and offers audio-replace plus lipsync repair passes when speech drifts. - Use Case: A creator says "put me on a beach at sunset" for a 12-second talking clip; the Skill normalizes the video, edits the face frame, waits for approval, generates the restyled video, and QA-checks duration, speech, and identity before returning the result. ## Quick Start Use the fix-my-look skill to change the background of my video at ~/Downloads/clip.mp4 to a neon-lit city at night while keeping my face and voice exactly the same.
