LLM Transcription Skill

Audio to markdown transcription converts recorded meetings into structured text using the Whisper API, including speaker identification and clear section breaks. It processes audio files like interviews or lectures into searchable, editable markdown documents.

Converting images to markdown uses ensemble, judge, and refinement steps to extract text from screenshots and diagrams while preserving structure. This approach ensures formatted markdown output that maintains the original document layout and graphical data.

Batch processing handles multiple files efficiently for large-scale audio and image transcription needs. It converts bulk visual and audio information into structured markdown text, making the content searchable and usable in downstream applications.

Image to text conversion supports multiple models from both OpenAI and Anthropic. This multi-model support allows flexible LLM selection during the ensemble, judge, and refinement phases to optimize markdown extraction from documents and graphics.

Transcribing a scanned PDF involves image-to-markdown conversion that preserves document structure and extracts data from embedded graphics. The process uses ensemble steps and judge logic to output reusable, formatted markdown text.

Using LLMs for transcription goes beyond standard OCR by preserving document structure and extracting data from graphics. The ensemble and refinement steps format the extracted content directly into markdown, providing searchable and editable text for downstream use.