All lessons

Lesson 9 of 16

Seedance 2.0 and 2.5, two dialects

Members

The two Seedance dialects that contradict each other, the face guard that refuses stills, both rule sets in full, two annotated pastes, the failure modes and the costs

26 min readUpdated Sep 3, 2026
  • Seedance 2.0
  • Seedance 2.5
  • Dreamina

Seedance generates audio and video together, takes images, clips and audio as references, and can edit and extend its own output. It is behind most of the recent RealityDrift finals. It also has two current models whose prompt rules contradict each other on timing, on sound notation and on negatives, and a paste written for one degrades silently on the other. Learn them as two dialects of one engine, built from the same master playbook in your prompt file and never from each other.

The official name of the flagship is plain "Dreamina Seedance 2.0". Resellers call it "2.0 Pro"; it is the same model, and its 720p list price is $0.15 per second.

In this lesson

  • Two doors
  • Running a take in Dreamina
  • The rule that applies to both: E005
  • Seedance 2.0: event order, symbols, a constraint tail
  • Seedance 2.5: timestamps, plain audio, positive statements
  • The raptor 2.0 storyboard, annotated

This lesson is part of the course. The Lab membership includes all of it, plus every new lesson and revision. The Previz Pack includes the lessons published at the time you buy.