7-Pose DSLR Character Sheet

date: '2026-05-01T03:30:22+08:00'
result: pass
prompt:
  text: >-
    DSLR Character Sheet on a white back drop wearing a white tea and a bomber jacket. With three
    face profiles (front, 45° and side) and four full body portrait profiles (front, 45°, side and
    back).
refs: []
provider:
  id: gpt_image_2
  relay: apimart
  config:
    aspect_ratio: '4:3'
    size: '4:3'
    'n': 1
output:
  path: ./dslr_7pose_charsheet_v1.png
  bytes: 915527
  wall_seconds: 44.6
  task_id: task_01KQFXWGHGSBBHXJZ48P2F9MME
script: experiments/dslr_charsheet_test/test_v1.py
cost_yuan: 0.5
notes: >-
  Stage 1 of insmind_com 2-stage pipeline (image → video). DSLR 7-pose character sheet · pure T2I ·
  no refs. Tests: (a) does gpt-image-2 honor 'three face profiles + four body profiles' = 7 panel
  layout · (b) does 'DSLR' aesthetic anchor pull toward photoreal lighting · (c) outfit lock (white
  tee + bomber jacket) across all 7 panels · (d) implicit identity consistency (no ref · model
  invents · should auto-lock). Verbatim contains typo 'white tea' (intended: 'white tee') · kept
  verbatim per discipline. Compare against existing multiref-5frame-charsheet (5 panel ·
  validated_2plus on kitsunebi).