# Fast Feature V2 Interval8 Experiment

- Episodes: 5
- Baseline extraction: 361.90s
- Fast extraction: 218.61s
- Wall-time reduction: 39.6%
- Baseline raw visual GT recall: 16/17
- Fast raw visual GT recall: 16/17

| Episode | Baseline | Fast | Faster | Face valid Δ | Face IoU | Baseline→Fast | Fast→Baseline | GT baseline | GT fast |
|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|
| P24/EP20 | 73.24s | 42.58s | 41.9% | +6.1pp | 0.9337 | 100.0% | 96.6% | 4/5 | 4/5 |
| P24/EP21 | 70.67s | 45.12s | 36.1% | +11.3pp | 0.9416 | 99.4% | 99.4% | 1/1 | 1/1 |
| P24/EP23 | 69.86s | 43.05s | 38.4% | +8.4pp | 0.9361 | 100.0% | 97.9% | 3/3 | 3/3 |
| P24/EP25 | 80.80s | 51.45s | 36.3% | +12.5pp | 0.9295 | 98.9% | 93.8% | 1/1 | 1/1 |
| P26/EP16 | 67.33s | 36.41s | 45.9% | +9.6pp | 0.9471 | 100.0% | 97.6% | 7/7 | 7/7 |

## Family overlap

| Family | Baseline→Fast | Fast→Baseline |
|---|---:|---:|
| black | 5/5 (100.0%) | 5/5 (100.0%) |
| blur | 213/213 (100.0%) | 235/242 (97.1%) |
| deformation | 281/284 (98.9%) | 277/297 (93.3%) |
| flicker | 246/246 (100.0%) | 246/246 (100.0%) |
| judder | 115/115 (100.0%) | 115/115 (100.0%) |

## Interpretation

- Fast extraction preserves the full per-frame timeline; no global FPS reduction is used.
- Black, flicker, and judder candidates are unchanged within the 0.5s comparison tolerance.
- Differences are concentrated in marginal deformation and blur candidates; raw visual GT recall did not decrease in this five-episode sample.
- Fast mode matches baseline raw GT coverage in this sample; this does not establish post-VLM equivalence.

Candidate overlap uses same-family intervals with 0.5s tolerance. GT recall is raw detector coverage with 1.0s tolerance and excludes lip-sync/out-of-scope rows. This experiment does not measure post-VLM ranking precision.