Seedance 2.5, 480p. Prompt 👇

Vertical 9:16 medium-shot photograph of a middle-aged man with graying hair and a full graying beard, wearing a dark navy blue button-down shirt with the sleeves relaxed, standing behind a small light-colored light-wood table in a bright classroom or small office. On the table in front of him sits a large round cream-colored cushion or bolster. He holds a brown leather belt looped around the cushion with both hands, fingers positioned near the metal buckle, demonstrating the grip to the room rather than looking at the camera; calm, instructional expression, mid-explanation. Behind him, a large white whiteboard covers the wall, carrying a hand-drawn marker diagram of a human torso and neck with arrows tracing flow lines upward, the words "LYMPHATIC SYSTEM" written across the top in the presenter's own handwriting. To the right of the diagram, two hand-written bulleted lists in the same marker: the first reads "DRAINAGE", "CIRCULATION", "LESS STAGNATION", "MORE ENERGY"; the second under the heading "KEYS:" reads "MOVEMENT", "HYDRATION", "BREATHING", "LYMPH SUPPORT". Left of the man a black flat-screen television is partially visible mounted on the wall. In the bottom-left foreground, the out-of-focus back of a seated audience member's head and shoulders enters the frame, shot slightly wider so it reads as a real room. Bright, even, neutral lighting with no dramatic shadows; documentary educational tone, realistic phone-camera or webcam-still quality, natural colors, sharp focus on the presenter, no text overlays, no captions, no watermarks. But that can only take you so far... (first frame btw) A good prompt isn't what's allowed it to be live for 26 days SomeONE had to generate each clip, make sure it's even usable, edit them together, add transitions, add sound fx, iterate, regenerate,e tc AI makes it cheap & easy but humans make it convert.

Work with me:

You could use Omni flash and get the same results or better, same with seedance 2.0 have you experimented those hey bro?

I’ve played around with both but found that 2.5 gives me the best results! Might give the same prompt to all three and see how they perform now

That's the thing, same prompt doesn't work for all models, you can say "make a cat", every model will make it differently, thus system prompts are something you should experiment with, whatever works for omni flash is different from what works for seedance 2.5, like if your prompt is very long, omni might not able to make a good video but if it's straight forward - it might do better than seed 2.5 and I pretty much know brands spending millions on ads monthly using only omni flash

Nice - how many SD2.5 clips did it take to make this?

seedance is so bad haha and takes 5 minutes for one video to generate