# Guide: Consistency

> Keep a character or style stable across assets.

For a character, mascot, product or room that must look the same across many
assets. All rules `[measured]` on DeepInfra FLUX-2 / nano-banana unless marked.

1. **Views generated independently produce a different subject each time.**
   Use the two-phase flow: generate N anchor candidates → a human picks one →
   every other view is generated conditioned on the picked anchor. Skipping
   the pick gives "three views of three different people".
2. **One sample per view drifts too.** Non-anchor views need N candidates
   (seed+k), a per-view pick and an acceptance check.
3. **The appearance string must be complete**: full outfit including lower
   body and footwear, exact colours. Anything unspecified is re-rolled per
   render. Inject it verbatim into every prompt; never re-describe per image.
4. **Lock bodies, not just faces.** Hands and legs give AI art away; a
   dedicated hands plate is the targeted defence `[inference]`.
5. **Identity first, set last** in the reference order, and cap references
   yourself: the provider drops extras from the end without error.
6. **Round-robin across subjects** when several need references, reserving
   one slot for the background plate; otherwise one subject eats the budget.
7. **Single-reference models take a collage**: 2, 4, 6 subjects side by side
   in one image were all recognisable; 8 gave 7 of 8. Poses copy the lineup,
   so collages suit wide shots, not close-ups.
8. **Background plates are copied near-literally.** One plate per camera
   position; a plate from another angle puts the wrong room behind the shot.
9. **Reference images must be caption-free**: text in a reference teaches the
   model to draw text.
10. **Test honouring before trusting**: generate fixtures, verify each
    against its own description with a judge, then render a scene whose
    prompt never describes the fixtures, with a no-reference control. A
    returned image proves nothing; only an output showing the reference's
    distinctive features does.
11. **Promote approved outputs to references** sparingly: human-approved
    only, never more than one generation removed from the locked set, or
    drift compounds `[inference]`.
12. **Editing a locked identity invalidates everything after it.** Warn at
    the point of edit and regenerate the reference set.
13. **Deterministic seeds** (base + index) make reruns repeat; identical
    payloads are cache hits, so "variants" must differ by seed.