One white-void transformation in three look targets: same beats, different characters

Three Seedance 2.5 runs of 25 August 2026 (UTC) use different look targets (a premium 3D animation look with one linked image; a midpoint between animation and photography; and a photographed-performer look) for a similar white-void transformation. H01 viewed 12 timestamped frames of each. Same in the sampled frames: a wide white frame with a tiny dark shape, a crouching figure, a rise, a close-up of an eye, an extreme close-up of a crystal in the chest and a silhouette leaving frame; two of the three end on a blank or near-blank frame. Different: the linked-image run has black hair and a ragged torn outfit; the other two show red hair and spiked dark armour. The photographed-performer target does not read clearly as a photographed performer at the sampled size and looks close to the midpoint target. Reading: in these three single runs (no seed set) the look target changed the character rendering less than the beat list did; this describes this set, not a rule. These are retrospective observations of founder tests, not a benchmark or product validation. Audio was not listened to.

{
  "citations": [
    {
      "note": "https://veyrul.com/experiments/exp_01M64VBFT7V2RAVYGFYNSMGQ3X (candidate; Main resolves to a record ID)"
    },
    {
      "note": "https://veyrul.com/experiments/exp_01M64VBKD0NH51QKJHYFTY18M3 (candidate; Main resolves to a record ID)"
    },
    {
      "note": "https://veyrul.com/experiments/exp_01M64VBHD5W39R96WCBJ1D9FHS (candidate; Main resolves to a record ID)"
    }
  ]
}