Propose→review→apply mechanical-residue scrub for session_doc narration prose. A deterministic regex pass surfaces candidate residue (raw numbers on DC/AC/HP/damage/healing/feet/rounds/initiative, out-of-fiction table-speak, real player names as speakers) —…
Quote-level consistency check on scene_extractions_new/ — flags VTT transcription errors, garbled phrases, and clarity issues in verbatim quote blocks, then proposes corrections for user approval. Invoke as /session-summary-consistency [session-dir].
Run consistency checks at every LLM-pipeline boundary (gm-assist → session-summary → scene extractions → narration) with a human-review gate between stages. Use when the user invokes /staged-consistency [session-dir] and wants the multi-stage check rather…
Clean up Otter/Zoom VTT transcripts for a D&D campaign — applies the known-misspellings glossary and prompts the user about unrecognised proper nouns. Invoke as /vtt-spell-pass [vtt-path].
Put speaker names on a session recording when the transcripts carry no usable attribution — the single-room case, where everyone shared one microphone so Zoom labelled all 866 cues with the host's name, and the editor's diarization produced six anonymous…
Rebuild a session transcript that has correct timings OR correct speakers but not both — the split-brain case where `*.speakers.vtt` carries fabricated timestamps and Zoom's live text export flips speaker mid-sentence, shredding one person's sentence across…
Quote-level consistency check on scene_extractions_new/ or scene_extractions/ for VTT transcription errors, garbled phrases, proper-noun misspellings, and clarity issues in verbatim quote blocks. Use when the user asks for /session-summary-consistency…
Render corrected scene-extraction quote blocks into readable, in-voice prose in a derived scene_extractions_smoothed/ layer. Use when the user asks for /voice-smooth [session-dir] or wants voice-aware smoothing after session-summary-consistency.