Try it

  1. Number the settings and outputs of candidates you might accept.
  2. Change only one condition, such as camera movement, in the next prototype.
  3. Compare using the same criteria and briefly record why you accepted a result.
Example prompt
Version A: Static camera.
Version B: A slow push-in.
Keep the subject, input image, lighting and other settings unchanged.

Requirements and considerations

If the model supports specifying a seed, keeping it fixed helps comparison. A seed alone does not guarantee character consistency or reproducibility across environments.

Official guides and references