What problem does it solve? Creating video that keeps a character, product, or brand style consistent across shots is hard with prompt-only generation. This Skill produces 4-30 second 1080p clips from ByteDance Seedance 2.5 Reference to Video on RunComfy, locking identity and style to your reference images, clips, and audio instead of prose descriptions. ## Core Features & Use Cases - Reference-conditioned generation: Feed up to 9 reference images, 1-3 reference video clips, and 3 audio files into one call to anchor identity, camera motion, and mood. - Native synchronized audio: The generate_audio option produces speech, SFX, and music in the same pass at delivery-resolution 1080p. - Cost-controlled workflow: Documents the counted-seconds billing model ($0.53 per second of reference plus output duration) and a 480p draft-then-deliver pattern for cheap iteration. - Use Case: A marketer needs a consistent-character brand clip. They supply three identity stills and a 4-second handheld camera plate, and receive an 8-second 1080p vertical video with ambience audio. ## Quick Start Ask the agent to generate a Seedance 2.5 reference-to-video clip using your reference image and clip URLs with a descriptive prompt, for example: "Use Seedance 2.5 reference to video to make an 8-second 9:16 clip of the woman in these reference images walking toward camera with this handheld follow plate."
