Seedance

Build the scene as an image first, then animate it with Seedance 2. Create the scene in an image model such as GPT Image 2 or Nano Banana until it looks authentic and not plastic. Then use that image as the reference in a video model such as Seedance 2 or Veo 3.

Images are cheaper to iterate on, while prompting video directly burns credits. Seedance 2 is probably trained on TikTok data, so it suits authentic UGC-style video. Higgsfield gives access to many models under one subscription. Quote: "it's going to be much easier to get it right with an image model and it's much cheaper."

Superwall▶ 36:00