Surface Generate errors in the guide panel; raise LLM max_tokens to 4096

Add a Capture button directly under the player

Reorder UI: extract frames (step 4) now follows transcribe + generate

Auto-extract: generate the LLM step outline first, anchor frames to step times

Auto-extract transcribes first when no transcript exists

Make auto-frame extraction narration-accurate

Fix lightbox on Spaces: inject via demo.load(js=) not launch(head=)

Enlarge selected frame in a full-screen lightbox

Add per-image delete from the frames pool

DocuMaker: video to step-by-step DOCX guide (Whisper + HF LLM + BLIP)