Seedance 2 is a multi-modal AI video generation model that enables precise control over style, subject, motion, camera, and audio through reference-anything technology, available on the Seedance Pro web platform.
What is Seedance 2?
Seedance 2 is a multi-modal video generation model developed by Seedance Pro that accepts text, image, video, and audio inputs and produces synchronized video with optional audio output. It runs entirely in the browser on seedance-pro.org, supporting up to 9 images, 3 videos (total under 15 seconds), and 3 audio files (total under 15 seconds) per generation, with output durations from 4 to 15 seconds and resolutions up to 2K.
Key Features
- Reference-Anything Control — Assign any input as a reference for style, subject, motion, camera, or audio using
@tags in the prompt. You can mix multiple references to precisely dictate every aspect of the output. - Multi-Modal Inputs — Combine up to 12 files (images, videos, audio) with a text prompt. Supports text-to-video, image-to-video, and video-to-video workflows in a single generation.
- Motion & Camera Replication — Replicate complex camera movements (Hitchcock zoom, orbit shots, tracking) and character actions from reference videos, enabling consistent multi-shot sequences.
- Video Extension & Editing — Extend existing clips by specifying duration (up to 15s total) or replace specific elements (characters, props, backgrounds) via text instructions without re-recording.
- Built-In Audio Generation — Generate dialogue, ambience, and sound effects synced to the video content and rhythm. Short audio cues improve lip-sync accuracy.
- Omni-Reference Mode — For multi-asset mixing, use Omni-Reference to assign distinct roles to each input (first frame, style, motion, camera, audio) for maximal control.
- Duration & Resolution Flexibility — Output lengths adjustable from 4 to 15 seconds; supports 720p default and up to 2K resolution on supported configurations. Aspect ratios include 16:9 and others.
