Deforum's strength_schedule sets how much the previous frame constrains the next and also fixes the effective step count

The strength_schedule parameter is a global per-frame control balancing how much the previous frame is retained versus resampled from the diffusion model. High strength (e.g. 0.9) makes each frame closely resemble the last — smooth, slow evolution but risking ‘stagnation’ where the prompt cannot drive change. Low strength (e.g. 0.3) frees the model to deviate, risking flicker or loss of coherence. It also sets the effective diffusion steps via steps - (strength * steps), so higher strength means fewer steps actually run per frame. This dual role means adjusting strength affects both visual smoothness and computation time.

Examples

strength_schedule 0:(0.65) runs 35% of steps per frame — moderate coherence. Lowering to 0:(0.4) during an expressive section lets prompt changes land faster but may flicker.

Assessment

If you set strength_schedule to 0.9 on a 20-step diffusion model, how many diffusion steps actually run per frame? What visual trade-off does this create compared to strength 0.5?