Update characters documentation, taxonomy, measured progress and validation

CLASSES.md

@@ -41,22 +41,22 @@
 | Outfit style | People | Images |
 |---|---:|---:|
 | `baby_everyday` | 16 | 160 |
-| `baby_layers` | 14 | 140 |
-| `baby_sporty` | 14 | 140 |
-| `casual_layers` | 38 | 380 |
-| `child_creative` | 20 | 200 |
-| `child_everyday` | 24 | 240 |
-| `child_formal` | 20 | 200 |
-| `child_outdoors` | 24 | 240 |
-| `child_school` | 24 | 240 |
-| `child_sporty` | 24 | 240 |
-| `crafts` | 28 | 280 |
-| `everyday_dress` | 14 | 140 |
-| `formal` | 31 | 310 |
-| `formal_skirt` | 14 | 140 |
-| `outdoors` | 35 | 350 |
-| `rock_inspired` | 28 | 280 |
-| `simple_everyday` | 28 | 280 |
+| `baby_layers` | 10 | 100 |
+| `baby_sporty` | 18 | 180 |
+| `casual_layers` | 31 | 310 |
+| `child_creative` | 30 | 300 |
+| `child_everyday` | 21 | 210 |
+| `child_formal` | 21 | 210 |
+| `child_outdoors` | 29 | 290 |
+| `child_school` | 17 | 170 |
+| `child_sporty` | 18 | 180 |
+| `crafts` | 25 | 250 |
+| `everyday_dress` | 17 | 170 |
+| `formal` | 27 | 270 |
+| `formal_skirt` | 6 | 60 |
+| `outdoors` | 36 | 360 |
+| `rock_inspired` | 40 | 400 |
+| `simple_everyday` | 32 | 320 |
 | `smart_casual` | 35 | 350 |
-| `sporty` | 38 | 380 |
-| `workwear` | 31 | 310 |
+| `sporty` | 43 | 430 |
+| `workwear` | 28 | 280 |

README.md

@@ -27,7 +27,7 @@ configs:
 
 **500 fictional people ×10 natural facial expressions =5,000 portrait images**, each832×1,248 pixels. The selected aesthetic is the **ghibli-warm-painted** direction: restrained facial proportions, delicate drawn lines, warm painted shading and a solid chroma-green backdrop. Generation uses the **undistilled9B base**, the **rank64 character LoRA**, **50 steps**, real **CFG4**, and **BF16**, with no quantization or distilled checkpoint.
 
-**Production snapshot:7/500 complete people;70/5,000 images uploaded.** Pilot and production validation are recorded in the linked metadata. Counts describe the snapshot; incremental shard commits can be newer.
+**Production snapshot:10/500 complete people;70/5,000 images uploaded.** Pilot and production validation are recorded in the linked metadata. Counts describe the snapshot; incremental shard commits can be newer.
 
 The [base model](https://huggingface.co/black-forest-labs/FLUX.2-klein-base-9B) and [character LoRA with examples and instructions](https://huggingface.co/laion/FLUX.2-klein-base-9B-vivarium-loras) are linked separately. The compatible [background dataset](https://huggingface.co/datasets/laion/Anim-E-Ghibli-Backgrounds-FLUX-9B) contains wide warm-painted environments for these foreground characters.
 
@@ -83,6 +83,8 @@ The full [class list](CLASSES.md), [taxonomy](metadata/taxonomy.json), [exact al
 
 All expressions request mild natural facial changes, with clearly specified mouth, brow and eyelid cues, rather than exaggerated caricatures. Joyful uses a natural open smile; grateful uses a softer appreciative smile. Pain means mild discomfort with no visible injury. Playful means an innocent lopsided smile. Fully clothed age-appropriate everyday outfits are used for all ages, including children and adolescents. Outfit styles range across casual layers, sportswear, formal clothing, practical workwear and rock-inspired denim; clothing avoids green and teal to support later chroma-keying.
 
+Clothing styles and colors use independent identity-hashed allocations, so the even/odd demographic assignment does not determine a clothing family. An early allocation refinement preserved the24 already started people and reassigned only the476 queued people; [the audit](metadata/outfit-allocation-refinement.json) records those changes. The published full person plan is authoritative for reproduction, including those retained initial outfits.
+
 The neutral image is generated first. **Each of the nine other images uses that same neutral image as its visual reference**, resized to416×624 using PIL BICUBIC, with the same seeded noise across the entire person group. Edits request the same identity, hair, skin, outfit, pose and framing and modify the face. The reference PNG's SHA256 and the reference sample ID are recorded. Reference conditioning encourages consistency; it is not a pixel-locking identity guarantee. Small pose, facial-detail or clothing changes can occur. The visual review records what was inspected rather than claiming all appearances were automatically verified.
 
 An initial pilot preserved identity but made facial expressions too similar. It was rejected before public ingestion. Subsequent pilots use focused face-edit prompts instead of repeating the full person description and remove the generic expression-suppression negative clause from edits. The accepted quality report, when available, records the actual inspected samples.

metadata/pilot-quality-passed.json

@@ -0,0 +1,16 @@
+{
+  "status": "passed",
+  "utc": "2026-10-06T15:47:26.833664+00:00",
+  "images_reviewed": 30,
+  "people_reviewed": [
+    "person-0002",
+    "person-0120",
+    "person-0253"
+  ],
+  "findings": "Focused edits preserve reference person,outfit and pose while giving visible mild facial changes",
+  "limitations": [
+    "Reference conditioning can slightly alter facial details and clothing folds",
+    "Visual assessment of rendered age and identity is subjective; demographic labels are intended prompt attributes",
+    "Mild fear,sadness and disgust can be visually close; labels record the requested expression,not a ground-truth expression classifier"
+  ]
+}

metadata/plan-token-validation.json

@@ -2,11 +2,11 @@
   "status": "passed",
   "persons": 500,
   "images": 5000,
-  "max_tokens": 232,
+  "max_tokens": 231,
   "min_tokens": 152,
   "truncated_prompts": 0,
   "unique_person_ids": 500,
   "unique_image_ids": 5000,
-  "plan_sha256": "0281a33dcec3a60dcf6a1f399162e21f9168f4b85188724e23fdb7afd5315f00",
+  "plan_sha256": "8f35392cc3c7bef3dde4262cc8ec4e400a18346b6253844455e90a5e05e98f0a",
   "tokenization": "Official pipeline chat template with enable_thinking=False; no truncation in validation"
 }

metadata/progress.json

@@ -1,18 +1,18 @@
 {
-  "utc": "2026-10-06T15:41:11.956309+00:00",
+  "utc": "2026-10-06T15:47:31.239912+00:00",
   "people": {
-    "done": 7,
-    "pilot": 3,
-    "queued": 483,
-    "running": 7
+    "done": 10,
+    "queued": 476,
+    "running": 14
   },
   "shards": {
+    "ready": 3,
     "uploaded": 7
   },
-  "images_in_closed_shards": 70,
-  "images_generated": 108,
+  "images_in_closed_shards": 100,
+  "images_generated": 205,
   "target_people": 500,
   "target_images": 5000,
   "images_uploaded": 70,
-  "buffer_free_gb": 302.112284672
+  "buffer_free_gb": 302.03314176
 }

metadata/provenance.json

@@ -23,7 +23,7 @@
   ],
   "reference": "The neutral image of the same generated fictional person, resized with PIL BICUBIC",
   "seed_method": "First32 bits of SHA256(vivarium-people-2026-10-06:<personality_id>),big-endian; same seed for all ten expressions",
-  "plan_sha256": "0281a33dcec3a60dcf6a1f399162e21f9168f4b85188724e23fdb7afd5315f00",
+  "plan_sha256": "8f35392cc3c7bef3dde4262cc8ec4e400a18346b6253844455e90a5e05e98f0a",
   "software": {
     "torch": "2.7.1+cu126",
     "transformers": "5.18.0",

metadata/quality-review.json

@@ -39,12 +39,53 @@
     "person-0133-sad",
     "person-0133-annoyed",
     "person-0133-pain",
-    "person-0133-disgust"
+    "person-0133-disgust",
+    "person-0120-sad",
+    "person-0253-sad",
+    "person-0002-afraid",
+    "person-0120-afraid",
+    "person-0253-afraid",
+    "person-0002-disgust",
+    "person-0120-disgust",
+    "person-0253-disgust",
+    "person-0002-pain",
+    "person-0120-pain",
+    "person-0253-pain",
+    "person-0002-annoyed",
+    "person-0120-annoyed",
+    "person-0253-annoyed",
+    "person-0002-grateful",
+    "person-0120-grateful",
+    "person-0253-grateful",
+    "person-0002-playful",
+    "person-0120-playful",
+    "person-0253-playful",
+    "person-0024-neutral",
+    "person-0024-joyful",
+    "person-0024-grateful",
+    "person-0024-playful",
+    "person-0024-angry",
+    "person-0024-afraid",
+    "person-0024-sad",
+    "person-0024-annoyed",
+    "person-0024-pain",
+    "person-0024-disgust",
+    "person-0089-neutral",
+    "person-0089-joyful",
+    "person-0089-grateful",
+    "person-0089-playful",
+    "person-0089-angry",
+    "person-0089-afraid",
+    "person-0089-sad",
+    "person-0089-annoyed",
+    "person-0089-pain",
+    "person-0089-disgust"
   ],
   "findings": [
     "Focused reference edit prompts produce clear natural neutral/joyful/angry differences for baby,schoolgirl and17-year-old girl while maintaining closely matched identity,hair,outfit and pose. Baby sadness was inspected separately when present.",
     "All ten expressions visually inspected for production baby person0001 and preschool girl person0068. Identity,hair,outfit,pose and greenscreen remain closely matched. Joyful,grateful,playful,angry and annoyance are clearly differentiated; several mild negative emotions can overlap. No extreme caricatures or visible injuries.",
-    "The complete older-schoolboy series person0133 also retains near-identical clothes and pose with restrained face proportions and ten mild requested expressions."
+    "The complete older-schoolboy series person0133 also retains near-identical clothes and pose with restrained face proportions and ten mild requested expressions.",
+    "All30 final focused-edit pilot images visually inspected across baby,schoolchild and older teen. Reference identity and outfit remain closely matched; joyful,grateful,playful,anger,annoyance and discomfort visible. Mild fear/sadness/disgust overlap. Production infant person0024 and schoolboy person0089 also inspected in all ten expressions."
   ],
   "status": "production pilot passed",
   "limitations": [
@@ -52,5 +93,5 @@
     "Visual assessment of rendered age and identity is subjective; demographic labels are intended prompt attributes",
     "Mild fear,sadness and disgust can be visually close; labels record the requested expression,not a ground-truth expression classifier"
   ],
-  "utc": "2026-10-06T15:38:26.256414+00:00"
+  "utc": "2026-10-06T15:47:26.833664+00:00"
 }

metadata/shards.json

@@ -1,4 +1,37 @@
 [
+  {
+    "name": "pilot-person-0002.tar",
+    "worker": 7,
+    "count": 10,
+    "bytes": 12503040,
+    "sha256": "c5e5a64531a22cddb2155ceafe7577ecc6d54c599fbad01ebf82d1970d9dde42",
+    "status": "ready",
+    "revision": null,
+    "created": "2026-10-06T15:47:29.374570+00:00",
+    "uploaded": null
+  },
+  {
+    "name": "pilot-person-0120.tar",
+    "worker": 7,
+    "count": 10,
+    "bytes": 10936320,
+    "sha256": "42dbc400abd92e0371195dc787b94d74c1d1bb85b439a3bd91a85f85652396bb",
+    "status": "ready",
+    "revision": null,
+    "created": "2026-10-06T15:47:29.405748+00:00",
+    "uploaded": null
+  },
+  {
+    "name": "pilot-person-0253.tar",
+    "worker": 7,
+    "count": 10,
+    "bytes": 11120640,
+    "sha256": "d46ce2eddc9fb2790647de358aafa345cea007b18dc0c033bbbdfa1b2e921986",
+    "status": "ready",
+    "revision": null,
+    "created": "2026-10-06T15:47:29.436613+00:00",
+    "uploaded": null
+  },
   {
     "name": "train-g0-00000.tar",
     "worker": 0,

tools/plan_character_production.py

@@ -128,7 +128,11 @@ def main():
                 hair=f'{hair_color} {style}'
                 if group=='baby':hair=f'soft sparse {hair_color} baby hair'
                 body=rng.choice(['a slim build','an average build','a sturdy build','a compact build']) if age>=16 else 'a natural age-appropriate build'
-                style_id,clothes=outfit(age,gender,rng,j)
+                # Keep the established age/hair/skin RNG sequence while making
+                # clothes independent of the even/odd ethnicity-cell index.
+                rng.choice(COLORS)
+                outfit_rng=random.Random(seed(ident+'-outfit-color'))
+                style_id,clothes=outfit(age,gender,outfit_rng,seed(ident+'-outfit-style'))
                 p=dict(personality_id=ident,character_id=ident,age_group=group,age_years=age,age_months=months,
                     gender=gender,ethnicity=eth[0],ethnicity_prompt_label=eth[1],skin_tone=rng.choice(eth[2]),hair=hair,
                     build=body,outfit_id='outfit-01',outfit_style=style_id,outfit=clothes,seed=seed(ident),default_emotion='neutral')