Kling 2.6 vs Seedance 2 for a cinematic book trailer — which do you reach for?

[pelegs202-design]

Hey — building a 60s cinematic book trailer through OpenMontage (Concept: intimate techno-thriller, 14 video-gen shots + 2 Remotion composites, 2.39:1 letterbox, moody_dark grade, ~$2 budget via fal.ai).

The video-gen-prompting.md skill has a dedicated Kling 2.6 guide (4-part structure, ++emphasis++ syntax) but no Seedance guide, so my prompts are calibrated to Kling by default.

Looking at the registry, seedance_video's declared strengths look like a strong fit for this project:

• native synchronized audio per clip (hazard ticks, key clacks, paper)
• director-level camera control + in-clip multi-shot
• high-fidelity motion with real-world physics
• lip-sync from quoted dialogue (I have ~10 VO lines)

Shot mix is ~10 intimate faces/inserts (Kling feels safe) and ~4 surreal/physics beats — desaturation pullbacks, headline morphs, stopped-car wide shots — where Seedance looks stronger on paper.

Two questions:

1. For a moody short-form cinematic piece like this, would you personally reach for Seedance 2 over Kling 2.6, or split them?
2. Any prompt-structure gotchas for Seedance that aren'''t captured in the generic video-gen-prompting.md yet? (Happy to PR a seedance-prompting.md if you have notes.)

Thanks for building this — the scored selector + Layer 3 skills architecture is genuinely excellent.

[calesthio]

Seedance2.0 is definitely the superior model in every aspect and changes to reflect that are now done. For fal.ai, Runway, Higsfield etc Seedance 2.0 would be the model of choice and the agent would automatically bubble it up when used