Seedance

Long-form single-take video: seamless continuous generation of 10+ minutes, with internal tests reaching 18 minutes without collapse. Narrative memory chain: retains plot beats, character personalities, and settings to support multi-act storytelling with suspense and twists. End-to-end video plus audio: generates video and audio together in one pass, removing separate dubbing and post-sync workflows. Native multilingual emotional dubbing: outputs tight lip-synced speech in Chinese, English, Japanese, Korean, with breathing, crying, and laughing matched to emotion. Hollywood-grade control plus cost reduction: supports storyboard/script input and real-time director commands with pro color grading looks, while cutting compute cost to 1/8th of Seedance 2.0 per cinematic minute.

Release timing: Seedance 2.0 was pre-released on Sunday, February 8, 2026 (reported February 10, 2026). Multimodal creation: It combines text, images, video, and audio to generate “cinematic” videos. More controllable output: ByteDance says it offers enhanced control via natural-language prompts (aimed at more directed filmmaking-style results). Voice-from-face capability (then paused): It reportedly could generate a person’s voice characteristics from facial images, and that feature was urgently suspended over risks. Creative control focus: The guide you linked emphasizes using multimodal inputs for precise control over motion, style, and storytelling (how-to claim, not an official spec).

Initial release of Seedance.