AI视频制作工作流分享

工作流程： - 首先编写完整的60秒序列。 - 将其拆分为两个相连的30秒部分。 - 为每个部分创建简单的12帧故事板。 - 锁定骑手、自行车、山地地形和下坡方向。 - 在Dreamina中：AI视频 → Dreamina Seedance 2.5 → Omni Reference。 - 添加故事板 + 角色 + 自行车参考。 - 每个镜头都使用真实的自行车物理特性进行提示：重量转移、制动、悬挂、着陆和恢复。 - 生成第1部分，然后从其最终状态继续第2部分。 最大的优势是控制。Seedance 2.5支持30秒生成和多达50个多模态参考，因此您可以使用实际制作资产来指导它，而不是依赖单个提示。 . . . [回复]

WORKFLOW: - Wrote the full 60-sec sequence first. - Split it into two connected 30-sec parts. - Created a simple 12-frame storyboard for each part. - Locked the rider, bike, mountain geography and downhill direction. - In Dreamina: AI Video → Dreamina Seedance 2.5 → Omni Reference. - Added storyboard + character + bike references. - Prompted each shot with real bike physics: weight transfer, braking, suspension, landing and recovery. - Generated Part 1, then continued Part 2 from its final state. The biggest advantage is control. Seedance 2.5 supports 30-sec generations and up to 50 multimodal references, so you can guide it with actual production assets instead of relying on one prompt. . . .