Every Seedance 2.0 post, filterable by category, sortable by date, searchable by keyword.
Two Seedance 2.0 filters block a lot of first tries: the face guardrail on real-person references, and the copyright guardrail on trademarked characters. Both live at the model layer.
Standard T2V is $0.3034/sec, Fast is $0.2419/sec for roughly 20% off. Here is the token formula, a 100 clip budget table, and when Fast loses quality.
Seedance 2.0 image-to-video holds wardrobe, face, and lighting for a full 15 seconds when you feed it the right still. The still does more work than the prompt.
Async queue pattern for Seedance 2.0: submit with webhooks, poll for status, track cost per job, and retry around the face filter without burning budget.
generate_audio is default true on Seedance 2.0. Here's when to keep it on, when to flip it off for a silent track, and what the lip sync tuning means for non English dialogue.
How to actually use the 9 image, 3 video, 3 audio reference surface on reference-to-video: file weighting, mixed vs image only, and when to skip audio refs entirely.
