| date | learning | source |
|---|---|---|
| 2026-10-06 | cartoon-h3 h3.py tells H3 'no on-camera dialogue, no lip-sync' in every mode; dialogue must be staged on listeners/off-screen until a native-speech mode ships. | scripts/cartoon_h3/h3.py video_prompt |
| 2026-10-06 | H3-max image-to-video with no audio input speaks a quoted line natively with lips (Fish approved 10-04 yapper). This is the route for a future on-camera dialogue mode. | project_dawn_yapper_2026-10-04 |
| 2026-10-06 | Dialogue pace keys that passed vo_check on 10-01: vo_tempo 1.08, gap_same 0.15, gap_turn 0.3, max_gap 0.22, seam_gap 0.25. | output/concept-packs/dawnbands-s2-attendance-office-h1-2026-10-01.md |
| 2026-10-06 | Recipe A in song form: vindication payoff $1,638 at ROAS 1.81; exhibit A 2.48; stake must land on mom and be caused by the mornings. | output/dawn-combo-map-2026-10-05/MAP.md |
| 2026-10-06 | Audit 10-06: an all-tagged pack is checked on vo_check DIALOGUE (overall >=2.9, line <=4.8 FAIL, turn gaps 0.30-0.55, same 0.12-0.30, dead air 0.6), not the 3.4 narration floor. VO lane ends 0.5 s after the last word and the 3 s end card overlays it, so the spoken CTA plays under the card. | scripts/cartoon_h3/vo_check.py run_check + DIALOGUE; assemble.py LAST_HOLD; endcard.py apply_endcard |
| 2026-10-06 | starpop articles: hook/hold/CTR loss-point read added to §11 (what each pattern means for THIS format's next variable). | ~/research/davidaistar/starpop/txt/ (slugs in references/sources.md) |
| 2026-10-06 | starpop articles: placement gate (Reels safe box y 14-65 %, Feed 4:5 crop) + Advantage+ enhancements-off/live-preview launch check. | ~/research/davidaistar/starpop/txt/ (slugs in references/sources.md) |
| 2026-10-06 | 10-06: splitter trap fixed (tagged lines staged as voice-over); gate = python3 -m scripts.cartoon_h3.check_plan --lint output/cartoon-h3/<slug> --pack output/concept-packs/<slug>.md = 0 BLOCK before approval. | scripts/cartoon_h3/storyboard.py, check_plan.py |
| 2026-10-06 | Built 10-06: run.py tiers (tiers.md), --tier-models, --draft 480P; h3.py prices fixed. Evals now expect the tiers.md check and the real flags. | scripts/cartoon_h3/run.py, h3.py, tiers.py (Builder B) |
| 2026-10-06 | sync-3 cartoon lip-sync proven 10-06 (Fish: 'lipsync was fine'): candidate route for on-camera dialogue once wired into cartoon-h3. | output/model-proofs-2026-10-06/README.md |
format-spoken-drama
Animated spoken drama ad, the "accused mom, vindicated" story (Dawn recipe A) told with voiced characters and a mom narrator instead of a song, 2-4 min.