What problem does it solve? Producing voiceover audio, caption timestamps, and transparent subject overlays for video compositions normally requires separate paid services and manual editing. This Skill provides three local CLI commands that generate these assets directly, with models downloaded and cached on first run. ## Core Features & Use Cases - Text-to-Speech: Generate narration locally with Kokoro-82M across 54 voices and 9 languages, with speed control and script-file input for long content. - Transcription: Produce word-level timestamped transcript.json from audio, video, SRT, VTT, or OpenAI API responses using Whisper models from tiny to large-v3. - Background Removal: Cut subjects out of video or images into VP9-alpha WebM, ProRes 4444 MOV, or PNG, with an optional inverse-alpha plate layer for text-behind-subject effects. - Use Case: Write a narration script, run npx hyperframes tts script.txt --voice af_heart --output narration.wav, then npx hyperframes transcribe narration.wav to get caption-ready word timestamps without any manual timing. ## Quick Start Generate a voiceover from my script file and transcribe it back into a word-level transcript.json for captions in my composition.
