ByteDance rolled out Dreamina Seedance 2.0 inside CapCut on March 26. The model generates video clips up to 15 seconds from text prompts or images across 6 aspect ratios. It's initially available in Brazil, Indonesia, Malaysia, Mexico, Philippines, Thailand, and Vietnam — notably absent from the US and Europe due to intellectual property concerns.
The model won't generate content from real faces and blocks unauthorized IP, with invisible watermarks applied to all outputs. Despite the restrictions, early demos show impressive quality — fluid motion, consistent characters, and coherent scene transitions.
Meanwhile, Lovart AI launched as a design agent platform with canvas-based editing, layer-level object manipulation, and multi-model integration pulling from GPT image-1, Flux Pro, and o3. Its standout feature lets you select an object in an image, drag it to a new position, and regenerate the surrounding scene — useful for creating start and end frames for video animation workflows.
The geographic restrictions on Seedance 2.0 highlight the growing fragmentation of AI tools along geopolitical lines. Users in blocked regions can access similar capabilities through other tools, but ByteDance's model quality suggests they'd prefer direct access if IP and regulatory issues were resolved.