Showcases

Vidu Q4 Preview Reference-to-Video: Character Consistency and Native Audio Demo

This entry is modeled on the reference-to-video input structure of Vidu Q4 Preview, demonstrating how to use a small number of reference images and a single voice reference to constrain character appearance and voice, then drive a short shot with native audio from a single prompt. Composition: Two character reference images lock the appearance and clothing of the protagonist and the antagonist respectively, and up to three audio references are used to fix the protagonist's voice (this work uses one Chinese dialogue voice reference). Following the schema convention, the prompt refers to the materials in the order "Reference Image 1 / Reference Image 2 / Reference Audio 1," constructing a brief confrontation scene: the protagonist's gaze and emotion build, the antagonist reacts with a turn of the body, the camera pushes in from a medium shot to a close-up, and ambient sound and dialogue are generated synchronously. Parameters: duration is set to 10 seconds, resolution to 1080p, aspect_ratio to 16:9, and audio is left on by default. Note that this model is a Preview version, and the official capabilities are explicitly stated to still be iterating; a single generation is capped at 16 seconds, so long shots need to be stitched in segments; reference images are limited to 15 and audio to 3 segments, and audio references must be 3–12 second MP3 files. Use cases: serialized short content requiring cross-shot consistency of character appearance and voice, previewing dramatic segments, multi-option comparison pitches (simply swap reference images or prompts to compare differences in performance and camera placement), and low-cost shot prototyping for independent creators and small studios.

Style Character Consistency · Native Audio · Conversation Scenarios

Model Vidu Q4 Preview

Reference inputs

reference_audio_urls
reference_image_urls
reference_image_urls

Try it yourself

Prompt

Try this model

Reference Image 1 defines the protagonist: a young woman in a worn olive field jacket, with short dark hair and a small scar above her left eyebrow. Reference Image 2 defines her opponent: a tall man in a charcoal coat, with a shaved head and rain-soaked shoulders. Reference Audio 1 provides the protagonist's voice.

Shot: at dusk, a rain-soaked concrete underpass, sodium lamps flickering, shallow puddles reflecting warm orange light. The opening is a medium shot with a slight low angle; the woman from Reference Image 1 steps out of the shadows, breathing quickly, her gaze fixed off-screen. In the voice of Reference Audio 1, she says in a low, restrained tone: "You shouldn't have come back." The man from Reference Image 2 slowly turns his head, his coat swaying, and answers without raising his voice. The camera slowly pushes in to a close-up of her face as she meets his gaze, her jaw tight, and blinks once; rain streaks across the lens, and fine debris drifts in the light. The camera stays handheld but steady, no cuts, one continuous camera move, natural motion blur, 35mm cinematic texture with delicate film grain.

Audio: synchronized dialogue, rain ambience, the low hum of distant traffic, footsteps on wet concrete, and a faint metallic clink at the moment of the cut.

Copy the full prompt and adapt it for your own generation.