fixing split

fixing splicer

Stitcher: absorb lone terminator cue from B when it completes A's sentence

Stitch sentences across scene boundaries so narration isn't split

Add skip_inverted_ocr flag to opt out of the inverted Tesseract pass

Drop garbled inverted-pass OCR lines that near-duplicate clean ones

Snap per-segment audio bounds to whole cues so clips don't end mid-sentence

Default to latest-in-time frame as the primary slide candidate

Anchor transcript alignment to video duration so guides cover full video

Keep review HTML re-fetchable and serialize audio_filename in readState

Validate segment filenames in /jobs/{id}/zip to block path traversal

Instructor filter: also filter alternates, promote first surviving alt

Instructor filter: prefer YuNet ONNX, Haar as fallback

Accessibility audit fixes: gating drafts, removing role conflicts, etc.

Filter: detect instructor frames the frontal-face cascade misses

Consolidate zip assets under static/

Three fixes: stale ends, /jobs cleanup window, audio toggle

Fix CLI audio output: inline for single, bundle for zip, schema round-trip

Sync deps, format validation, /jobs workdir cleanup, timing tests

Format on-screen OCR text into paragraphs, lists, and inline LaTeX

Tolerate blank max_frames on /jobs and /process

Live progress bar (web SSE + CLI rich) and rename images dir to 'static'

Phase D: --min-gap, --max-frames, --skip-ocr flag parity

Phase C: MathJax equations, key terms, and subtitle/module

Phase A: unified metadata schema + vgm CLI

Format narration into paragraphs

Restyle guide/review templates after photosynthesis prototype

Add browser-based review/edit step

Merge near-identical adjacent scenes after detection

Build initial VideoGuideMaker MVP