How video models work and which to use

Image-to-video uses your picture as the very first frame. Reference-to-video looks at your pictures while making every frame, so they guide the whole video.

Seedance 2.5 takes up to 30 reference pictures and makes videos up to 30 seconds. Omni takes only 3 pictures and stops at 10 seconds.

Slowing a video doesn't add frames: 24 slowed 3 times leaves only 8 a second. Ask for slow motion, or add frames first (interpolation).

For Hindi or other languages, the lesson picks Omni 1.1 Flash. Seedance's voices can sound off in any language other than English.

Seedance copies everything it sees in your reference pictures, including words. So give it pictures with no text on them.

On Omni, small words on signs and labels come out garbled. Add the text yourself in the edit.

Choose References, attach the sheets, describe the shot, set the length and shape, check the price on the button, then press Generate.

Make a cheap 480p draft on Seedance 2.0 Mini or MiniMax H3 Max Turbo. Fix the prompt there, then make the final just once.