Temporal artifacts—such as flickering, ghosting, or frame jitter—usually result from overly complex prompts, extreme motion scores, or insufficient inference steps. First, simplify your prompt by focusing on a single subject and clear action, then reduce the motion score to 20-40 for smoother results. If artifacts persist, increase the number of inference steps (denoising steps) from the default 28 to 40-50, which improves frame-to-frame consistency at the cost of slightly longer generation time. Additionally, refine your negative prompt to explicitly exclude 'temporal artifacts,' 'jitter,' and 'ghosting effects.' If you continue to experience issues, try
Seedance 2.0 Text to Video, which uses a different architecture optimized for temporal stability in character-driven scenes.