Nov 9 8 days ago 18:34 Drop 17 main images now rated 3; rename trigger lipb1te -> lpb1te Approved set is 94 pairs: 45 main, 32 legacy and 17 strong. All captions, metadata and manifest now use the V1 trigger lpb1te. dataset.json, README, summary, checksums and training-data.tar.gz regenerated. Nov 8 9 days ago 17:18 Add strong subset: 17 former legacy images with a stronger bite (batch-008) (part 2) Approved set is 111 pairs: 62 main, 32 legacy (rated 4) and 17 strong (rated 3). masks/strong are YuNet mouth-mask candidates, review not confirmed. dataset.json, README, manifest, summary, checksums and training-data.tar.gz regenerated. 17:18 Add strong subset: 17 former legacy images with a stronger bite (batch-008) Approved set is 111 pairs: 62 main, 32 legacy (rated 4) and 17 strong (rated 3). masks/strong are YuNet mouth-mask candidates, review not confirmed. dataset.json, README, manifest, summary, checksums and training-data.tar.gz regenerated. 06:45 Add reviewed mouth masks for the main subset (masks/main) 62 YuNet blend-ellipse masks for data/main, reviewed on overlay sheets; dataset.json gains main's mask block; checksums.json, README and training-data.tar.gz regenerated. Lets a trainer confine learning to the mouth so the LoRA stops learning the subjects' faces. Nov 4 13 days ago 05:02 Declare masks/legacy as non-image sidecar data. Add do_not_use_as_images so ingest does not treat mask files as photographs. 04:38 Drop 17 batch-006 images now rated 3 in Darktable. Approved set is 94 pairs: 62 main and 32 legacy. 04:28 Drop 14 images no longer rated 4. Approved set is 111 pairs: 62 main and 49 legacy. Nov 2 15 days ago 20:51 Describe the dataset files. Remove the workflow handoff from this repository. (part 2) 20:51 Describe the dataset files. Remove the workflow handoff from this repository. 20:45 Update WORKFLOW_HANDOFF.md 20:42 Restore 17 rating-4 main images dropped by the age-label caption filter. (part 2) 20:42 Restore 17 rating-4 main images dropped by the age-label caption filter. 20:35 Publish lip-biting V2 training set (part 2) 108 captioned pairs, legacy mouth masks, and the dataset.json training contract. 20:34 Publish lip-biting V2 training set 108 captioned pairs, legacy mouth masks, and the dataset.json training contract. 20:34 initial commit