What problem does it solve? Speech recognizers produce garbled, incomplete, or poorly formatted transcripts of lectures, especially for mixed English-Hebrew shiurim where Torah terms get mangled into transliterations or wrong English words. This Skill produces a faithful, complete transcript that downstream note-taking and study-material generation can rely on. ## Core Features & Use Cases - Full-coverage audio transcription: Transcribes referenced audio from the first spoken word to the last, with optional timestamps, speaker labels, and JSON schema-conformant output. - Transcript cleanup: Repairs raw speech-recognition text by removing filler, stutters, and timestamps, then rewrites it as continuous prose paragraphs without summarizing or adding content. - English-Hebrew shiur handling: Restores Hebrew script for Torah terms, source abbreviations (גמ׳, רש״י, תוס׳), and daf references that recognizers emit as transliterations, using a built-in mapping table of common misrecognitions. - Use Case: A student records a 40-minute Bava Metzia shiur and the recognizer outputs "the camera asks on toss those". The Skill restores it to "The גמ׳ asks on תוס׳" and returns a clean, complete transcript ready for outline notes and flashcards. ## Quick Start Transcribe the lecture audio at @recording.m4a into a complete transcript, or clean up the pasted raw transcript text into continuous prose paragraphs.
