| name | video-script-captions |
| description | Builds on-point video SRT from locked segment scripts plus Whisper timing only — never raw transcription text. Use for daily_single captions, final.srt, segment.srt, subtitles, transcript alignment, YouTube hook/overview/Let's get started structure, or when captions must match June roundup quality. |
Video script captions
Captions show script text, timed by Whisper. Same rule as June roundup (write_verses_srt): viewers read the authored script, not Whisper's guess.
Do NOT
- Put Whisper
segments[].text directly into SRT (misheard words, filler, blabber)
- Include editor labels (
Hook:, Beat N:, [VISUAL:…]) in caption text
- Regenerate captions from stale
merge/narration.json after script changes
- Skip the hook/overview/Let's get started block at the start
DO
| Pipeline | Command |
|---|
| daily_single | python -m praisonaippt.daily_single --project . build-captions |
| segment roundup | Per-segment segment.srt via write_verses_srt / write_cue_timings_srt; merge stitches offsets |
YouTube opening structure (June standard)
First SRT cues must follow this order:
- Hook sentence (attention)
- Overview sentence(s) — use comma-separated roll-call so montage timing can split one cue into N hero windows (June / Fable pattern)
Let's get started. — own cue; content beats follow after
June roundup hook (examples/videos/june-2026-ai-roundup/segments/00-hook/script.md):
Fifteen stories in this roundup: … Let's get started.
Single-topic hook — three cues minimum; overview cue should list topics separated by commas for hook montage. See .cursor/skills/daily-single-video/SKILL.md.
Script density
| Segment | Target |
|---|
| Hook | ~55–70 words: attention + overview + Let's get started |
| Topic / beat | ~2–4 plain-language sentences |
| Outro | June CTA (~25 words): like/subscribe — no mer.vin spoken |
Every sentence in script.md becomes one SRT cue. No extra words beyond the script.
daily_single workflow
python -m praisonaippt.daily_single --project examples/videos/<slug> build-captions
python -m praisonaippt.daily_single --project examples/videos/<slug> validate-display
Outputs:
segments/*/segment.srt — one cue per script sentencemerge/final.srt — stitched with merge/timeline.json offsets
Implementation: praisonaippt/daily_single/captions.py
segment-video-roundup (reference)
align-cues → cue_timings.json (Whisper anchors)yaml → verses with notes = script fragmentbuild → write_verses_srt(seg_dir, verses)merge → offset stitch → merge/final-roundup.srt
See .cursor/skills/segment-video-roundup/SKILL.md Phase 5–6.
Gate before publish
Additional resources
