看到有人发布逼真的 Seedance 2.5 视频。如果你打算用 i2v 思路，下面的提示词公式是有效的：

Seen folks post realistic Seedance 2.5 videos. If you are planning an i2v approach, a prompt formula like this works: 

2x2 grid, 16:9, AI picks subjecst: V₁  DIRECT‑ADDRESS MID‑SENTENCE   the subject looks INTO the lens and is caught mid‑speech (mouth open mid‑word, animated       brows) as a vlogger talking to their audience; not a posed smile, not looking away.   ¬V₁ = posed/still / averted gaze / closed‑mouth model smile. V₂  HANDHELD SELFIE FRAMING   arm's‑length close framing with wide‑angle character (slight edge distortion, subject       off‑center, head near the top edge, background sweeping), slight roll/tilt, imperfect crop with foreground intrusion       (duvet/shoulder/hand).   ¬V₂ = tripod composed / telephoto / centered symmetric / clean crop. V₃  INFERRED VLOG SITUATION + BEAT   a recognizable vlog beat (morning‑in‑bed duvet wrap · kitchen coffee · car passenger ·       gym post‑set · walking street · unboxing desk · hostel bunk · late‑night desk) with the subject mid‑activity; the beat       implies what they're saying.   ¬V₃ = no situation / generic backdrop / no activity. V₄  LIVED‑IN ENVIRONMENT + AVAILABLE LIGHT   a real lived‑in space with mundane props (window+tree, radiator, kettle,       bedding; car interior; gym; kitchen clutter) lit only by available light (overcast window, practical, screen glow).       ¬V₄ = studio/set / flash / sterile. V₅  NATURAL LOOK + CONSUMER‑CAMERA TRUTH   natural skin/hair (no glam), casual wardrobe/wrap; consumer imaging (phone/       front‑cam dynamic range, slight noise, soft contrast, no cinematic grade).   ¬V₅ = glam retouch / cinematic grade / HDR / pro lighting. η   HONEST DETAIL   window catchlight, real fabric folds, hair flyaways, readable background. FORBIDDEN = posed smile · averted gaze · tripod compose · telephoto · studio light · glam retouch · cinematic grade · sterile set · no situation
…