transcript-fixer

Automatically fix speech recognition errors in meeting transcripts by applying dictionary rules and AI analysis to correct garbled text, homophones, and technical terms. This Skill processes your text output to significantly reduce manual correction labor.

AI analysis corrects homophones and garbled technical terms in ASR output by identifying speech-to-text mistakes and applying dictionary rules. It targets recognition errors specific to lecture, interview, and meeting transcriptions.

To start correcting transcription files, you need to initialize the environment using uv and Claude Code. These dependencies manage the required packages like httpx and aiofiles for automated text processing.

This speech-to-text correction approach learns from previous fixes by building a personalized correction database. It accumulates rules from each applied fix to continuously improve accuracy over time for future transcriptions.

This error correction process handles text produced by speech recognition, including meeting notes, lecture transcripts, and interview recordings. It applies to any ASR or STT text output containing recognition errors.