v1.0

Initial public release.

• Multi-model Whisper ensembling with LLM-based adjudication
• Critical text merging from 2–3+ transcript sources (Whisper, YouTube captions, external transcripts)
• Blind/anonymous source presentation to prevent provenance bias
• wdiff-based alignment for accurate chunking across sources of different lengths
• Structured transcript preservation (speaker labels, timestamps)
• Slide extraction and optional vision API analysis
• Make-style DAG pipeline with checkpoint resumption
• Cost estimation and local-only mode