Remember from the day when we could only "tell a story" using text and the first/last frames, we've always wanted to create a video model that truly understands your expression. Today, it's finally here!
Seedance 2.0 now supports four modalities of input: images, videos, audio, and text, offering richer expression and more controllable generation.
You can set the style of the scene with a single image, specify the character's actions and camera changes with a video, and then set the rhythm and atmosphere with a few seconds of audio... With the help of prompts, the creative process becomes more natural, more efficient, and more like a real "director."
The biggest highlight of this upgrade is the "reference ability":
- Reference images can accurately reproduce the composition and character details of the scene
- Reference videos support the replication of camera language, complex action rhythms, and creative special effects
- Videos support smooth extension and connection, allowing the generation of continuous shots based on user prompts, not just generating but also "continuing to shoot"
- Editing capabilities are also enhanced, supporting role replacement, deletion, and addition to existing videos